Journal:Best practice recommendations for the implementation of a digital pathology workflow in the anatomic pathology laboratory by the European Society of Digital and Integrative Pathology (ESDIP)
|Full article title||Best practice recommendations for the implementation of a digital pathology workflow in the anatomic pathology laboratory by the European Society of Digital and Integrative Pathology (ESDIP)|
|Author(s)||Fraggetta, Filippo; L'Imperio, Vincenzo; Ameisen, David; Carvalho, Rita; Leh, Sabine; Kiehl, Tim-Rasmus; Serbanescu, Mircea; Racoceanu, Daniel; Mea, Vincenzo D.; Polonia, Antonio; Zerbe, Norman; Eloy, Catarina|
|Author affiliation(s)||European Society of Digital and Integrative Pathology, Azienda Sanitaria Provinciale Di Catania, University of Milano-Bicocca, Imginit SAS, Charité – Universitätsmedizin Berlin, Haukeland University Hospital, University of Bergen, University of Medicine and Pharmacy of Craiova, Sorbonne Université, University of Udine|
|Primary contact||Email: celoy at ipatimup dot pt|
|Editors||Sergi, Consolato, M.|
|Volume and issue||11(11)|
|Distribution license||Creative Commons Attribution 4.0 International|
The interest in implementing digital pathology (DP) workflows to obtain whole slide image (WSI) files for diagnostic purposes has increased in the last few years. The increasing performance of technical components and the Food and Drug Administration (FDA) approval of systems for primary diagnosis led to increased interest in applying DP workflows. However, despite this revolutionary transition, real-world data suggest that a fully digital approach to histological workflow has been implemented in only a minority of pathology laboratories.
The objective of this study is to facilitate the implementation of DP workflows in pathology laboratories, helping those involved in this process of transformation with: (a) how to identify the scope and the boundaries of the DP transformation; (b) how to introduce automation to reduce errors; (c) how to introduce appropriate quality control to guarantee the safety of the process; and (d) addressing the hardware and software needed to implement DP systems inside the pathology laboratory. The European Society of Digital and Integrative Pathology (ESDIP) provided consensus-based recommendations developed through discussion among members of the broader scientific committee. The recommendations are thus based on the expertise of panel members and on the agreement obtained after virtual meetings. Prior to publication, the recommendations were reviewed by members of the ESDIP Board. The recommendations comprehensively cover every step of the implementation of a digital workflow in the anatomic pathology department, emphasizing the importance of interoperability, automation, and tracking of the entire process before the introduction of a scanning facility. Compared to the available national and international guidelines, the present document represents a practical, handy reference for the correct implementation of a digital workflow in Europe.
Keywords: digital pathology, anatomic pathology workflow, whole slide imaging, laboratory information system
The interest in implementing digital pathology (DP) workflows to obtain whole slide image (WSI) files for diagnostic purposes has increased in the last few years. This is in part due to the opportunities offered by WSI, e.g., telepathology and image analysis, including computational pathology tools based on artificial intelligence (AI) methods. The increasing performance of technical components and the Food and Drug Administration (FDA) approval of systems for primary diagnosis led to increased interest in applying DP workflows. Moreover, in the last few years, several studies evaluating performance demonstrated the non-inferiority of WSI compared to conventional light microscopy for primary histological diagnosis. This may help to alleviate concerns about the possible risk of DP-related diagnostic errors. Indeed, the restrictions suffered during the COVID-19 pandemic, the reduction in the number of pathologists, and the increase in workload, along with a rising number and complexity of clinical cases, also raised the interest in DP.
Several definitions for DP have been proposed so far, a common opinion being that DP encompasses the photographic documentation of the macroscopy of the specimens (“gross pathology”), the digitization of glass slides (virtual microscopy), and telepathology. By some definitions, DP involves merely the digitization of glass slides. In this study, “DP” is significantly distanced from the reductive paradigm of only glass slide digitization, moving towards a more integrative approach that comprises interventions in all stations of work in the pathology laboratory, introducing and supporting innovation. DP implicitly consists of all the associated technologies to allow improvements and innovations in workflow, including, for instance, laboratory information systems (LIS), digital dictation, dashboard and workflow management, electronic specimen labelling and tracking, and synoptic reporting tools.
The objective of this study is to facilitate the implementation of DP workflows in pathology laboratories, helping those involved in this process of transformation to: (a) identify the scope and the boundaries of the DP transformation; (b) introduce automation to reduce errors; (c) introduce appropriate quality control to guarantee the safety of the process; and (d) implement the hardware and software needed to implement DP systems inside the pathology laboratory. Since several recommendations and guidelines have already been proposed, primarily focusing on the validation of WSI for clinical purposes or on the technical environment, this paper mainly covers DP implementation and all the prerequisites for a pathology laboratory to change from an analog to a digital workflow. Considering all that has been reported about DP workflow implementation and its associated benefits, it is anticipated that this new methodology has many advantages that should be attractive and convenient for all pathology laboratories worldwide, independently of their dimension, workload, number of pathologists, or type of activity (e.g., academic/nonacademic, private/public).
So far, there are several possibilities to transit and to manage “images” in a digital workflow: an LIS-based approach, a scanner vendor approach, or an intermediate software approach (e.g., Linköping University). Independently of the type of strategy chosen, in order to switch towards a digital visualization of images (whether LIS-centric, vendor-based, or third-party software), the new system should be able to integrate every possible instrument (e.g., one or more scanners from the same or different vendors, with the possibility to manage different images from a variety of sources), preferably associated with a tracking system because of automation and innovation. The cost-effectiveness of DP has already been documented in implementation models that discuss the scope of investment, the potential return on investment, and cost-savings of DP, as well as any proposed income deriving from the adoption of WSIs. Moreover, the adequate adaptation of a routine clinical workflow can finally lead to an optimization of resources (e.g., space, time, personnel, and equipment). These are intended as recommendations and suggestions for the implementation of the full DP workflow in the routine clinical practice of anatomic pathology laboratories. The introduction of a DP workflow even allows the implementation of computational pathology tools such as AI.
The following sections explain, point-by-point, the steps needed for the progressive, secure, and efficient transition into a DP workflow. Regarding cytopathology, there are several barriers that still need to be overcome for routine cytopathology to go digital and support wider adoption and sustainability. Therefore, the present study mainly focuses on histopathology and the justification (Box 1) for its transition to a DP workflow.
|Box 1. The justification of transitioning histopathology workflows to digital pathology workflows.|
|1. Digital pathology is pathology, a holistic approach that comprehends interventions in all stations of work at the pathology laboratory, introducing innovation.|
2. Digital pathology is attractive and convenient for pathology laboratories worldwide.
Involvement of the team in the digital pathology transformation of the laboratory
The implementation of digital pathology requires a multidisciplinary approach from the very beginning. The leading team should involve in-house participants (e.g., pathologists, laboratory technicians, administrative staff) and the hospital’s IT and technical services. IT services might be organized in different ways depending on the size of the department and depending on local or national policies. For example, IT services may be provided by individuals, by a separate department, or by a subcontractor. The most important thing is that these groups work together and that they form a team.
Subsequently, close collaboration with companies providing the digital pathology system and the LIS will become necessary. Especially in larger departments, digital transformation will usually be organized as a project that includes a project manager, a steering group, and different working groups. There are several ways of introducing the topic and designing the appropriate options for the laboratory at hand, and it might be useful to visit pathology departments with digital workflows to learn from their successes and failures. There are a couple of papers that share such experiences and provide valuable information. Describing user scenarios is another method to both understand the needs of one’s own laboratory and communicate those needs to the IT and technical departments, as well as potential vendors. In addition, before starting a tender, it is helpful to gather information about suppliers and products.
To obtain a successful implementation of DP and to avoid deficiencies, the multidisciplinary team that is going to lead the “digital revolution” in each department should follow some crucial steps, as previously reported. In particular, for the correct and rapid implementation of DP in every department, it is advisable to create awareness and appropriate work conditions, incentivize participation, encourage communication among the team members, and monitor the outcomes of this revolution. This approach could help in facing the heterogeneous patterns of reactions that different actors of the team could express, including the “enthusiasts,” the “sceptics,” and the “undecided.” All the possible measures to increase the trust and involvement of pathologists should be applied to all staff members.
To establish a successful DP workflow, a thorough stakeholder analysis should be carried out, and a communication strategy should be established based on this analysis. The team must ensure that all internal stakeholders (including pathologists, laboratory personnel and administrative staff) are continuously informed from the beginning. In this setting, sharing the vision of DP with laboratory and administrative personnel, encouraging them to provide feedback, expressing potential concerns and suggestions (e.g., using frequent meetings on-site), and providing appropriate discussion during all phases of the deployment will facilitate a safe and effective implementation. The team must be aware that DP should be perceived as an integral part of the laboratory workflow rather than an “add-on.” The contingent situation due to the COVID-19 pandemic can be further leveraged to boost the implementation of DP in the laboratories, stressing the need to maintain pathology services by making it possible for pathologists to work from home.
Implementing DP as the standard laboratory practice also requires learning new technical skills to capture all the advantages of this technology. Just as significant as consulting the internal stakeholders is the involvement of IT services. IT will be crucial in many aspects of the project (e.g., handling LIS adaptations, integrations, storage, testing etc.). The involvement should start in the early phases of the transition. For example, consider conducting a laboratory office tour to establish communication with the other components of the project, using clear language. This gives your team the opportunity to understand what is expected and what is potentially achievable from your deployment, and what each professional group will be expected to contribute in terms of time and staff. Explain your ideas for future digital workflows and see what potential dependencies and solutions your IT colleagues can generate.
Optimization of resources in the DP workflow
In a fully digital laboratory, the processes and records are electronic file-based and the environment is paperless, with glass slides being substituted at the end of the workflow by WSIs. The optimization of resources—namely time, space, people and instruments—creates conditions for increased efficiency and, consequently, decreased costs. The LEAN approach represents a valuable strategy to optimize the workflow, leading to a more logical distribution of the spaces to minimize staff and specimen traffic inside the laboratory. It also allows for a more harmonic and well-planned articulation between human resources and available instruments, which results in time and cost-effectiveness. And, although it is not a strict prerequisite for adopting DP, it could further allow for better allocation of resources.
This reallocation could start from a more rational disposition of the spaces or offices inside the pathology laboratory. An inefficient arrangement of the physical spaces, typical of the old, “analog” workflow, can partly impair the smooth crosstalk among the different components of the process. Previous experiences in implementation models stress the need to analyze the pre-existing workflow before implementing DP. A careful analysis of the pre-existing analogic workflow before the transition should consider the flow of the specimens (workstation location) in the laboratory and time intervals (hands-on and waiting times) for each workstation, verifying the information technology support and establishment of adequate quality control checkpoints. The lack of structural organization of some pathology laboratories, including the physical placement of the different workstations, may contribute negatively to the desirable, efficient crosstalk between workstations. The reorganization of such a laboratory structure with the intent to decrease unnecessary movements of the staff, and time loss, can be useful for every laboratory, independently of DP implementation. For instance, the scanning workstation should be located near the staining and mounting instruments, accelerating the production line but far from the microtome area to avoid the interference of paraffin with the scanning mechanisms. After this retrospective analysis and reorganization of the structure, the optimal choices for the automation of each workstation must be made, namely by the introduction of a reliable tracking system, and different instruments would preferably work in a coordinated fashion, connected (mono- or bi-directionally) to the LIS or laboratory information management system (LIMS).
The role and potentialities of laboratory informatics resources
Independently of the system employed to manage the WSI, pathology laboratories mainly depend on a laboratory information systems (LIS) to support their operation and, ultimately, carry out their patient care mission. For these reasons, one of the crucial points is to ensure the full integration of the systems involved in the digital transition. Although many LISs have evolved with sophisticated and more user-friendly software over the past few decades, supporting a broader range of functions, many others have not evolved, thus preventing possible integration with other technologies deployed in pathology laboratories. A modern LIS plays different roles in all phases of patient testing, including specimen and test order entry, specimen processing, and specimen tracking. They track and organize the laboratory’s workflow, mainly through event logs and histology protocols. The maintenance of such logs can follow the default configurations of the system or can be customized by each laboratory to display the most useful information. A typical example of the system’s default configuration for a log (e.g., routine histology) includes accession number, timestamp, patient and specimen data, histology protocol(s) ordered, other stains ordered, and comments about the specimen or the request.
An LIS now incorporates multiple features that, until recently, were either unavailable or required a significant customization effort to be obtained. The Association for Pathology Informatics produced a comprehensive list of basic and advanced LIS features that may be used to evaluate LIS capabilities. Moreover, the next generation LIS should be able to link digital images to the respective cases appropriately. With the rising use of WSI for clinical purposes, a consensual increase in capabilities to connect and integrate WSI systems and LIS is to be expected (e.g., being able to open WSI from the LIS, log the viewed areas/magnification on all WSI, or even apply image analysis and store result data). Further advances in the development of the LIS are expected in the future, starting from the integration of more sophisticated tools to support data mining and the analysis of pathology and clinical data sets. The LIS may evolve into a multimodality “pathologist cockpit” that not only provides LIS functions but also displays pathology imaging and other medical imaging, supplies analytical tools, provides access to clinical data (e.g., through an electronic health record [EHR]) as well as other data sources. A more recent guideline paper underlined the importance of digital pathology interoperability, with a LIS being able to connect all the instruments present in the laboratory to support critical DP use cases. Moreover, increasing requests for molecular and genetic tests on pathology specimens (e.g., next-generation sequencing) impose further innovation on the LIS to integrate and optimize these data with the traditional pathology report for optimal patient management in an integrative model. Finally, the transition will allow information integration from grossing, enable collaborative work, and incorporate quality control results.
Automation of workflow and tracking system
Automation and using a robust tracking system can significantly reduce errors related to handwriting transcription and misspelling that can cause specimens to be dissociated from a particular patient (“mismatching”). Automation is a “strong recommendation” emanating from these recommendations, as it can benefit both pathology laboratories using DP and those using glass slides for diagnosis. Besides the introduction of a suitable LIS or LIMS that can help monitor the instruments’ performance connected to each specimen, further automation can be introduced in the lab's workstations. This includes tracking the reagents used and tracking all the staff that were at any point involved in specimen processing by differential log-ins or scanning of individual ID codes at all workstations. Yet the possible automation of workstations obviously depends on budget, existing instruments, and the experience of the technical staff. Devices such as a robotic stainer and a cover-slipper will bring consistent slide quality, avoiding frequent re-staining and ongoing readjustments to scanning protocols. The same is true for the automation of embedding and cutting processes, for which available systems on the market appear promising. However, these are not yet widely used in practice, perhaps in part due to cost.
The goals of any tracking system in the laboratory are to keep the specimen automatically, correctly, and permanently labelled during the time that it circulates in the laboratory. The identification of the specimen—using labels on the containers, printed in the cassettes/paraffin blocks, printed on the glass slides and present in the WSI files—is a best practice rule that is recommended to be adopted for the use of the WSI. In this setting, the perfect compatibility (interoperability) of the instrumentation used to label and to process the specimens within the anatomic pathology laboratory, and with the other laboratories in the same institution, is crucial to avoid possible issues (e.g., blurring or shading of the labels during subsequent processing of specimens/slides). The identifiers related to the specimen, of which there are usually several (see the next section), should be managed automatically and electronically connected to the patient’s LIS entry. The integration between the tracking system and LIS with an electronic interface between the LIS and the printers is essential to maintaining continuity of identification. The link established between the asset (tissue container/cassette/block) and the LIS will help reduce errors and can be achieved by printing different data types on the assets, such as barcodes or 2D (QR) codes. These can be linked to different types of data in the LIS. Eventually, other systems with code reader compatibility will be able to read them. The introduction of radiofrequency identification (RFID) technologies is also a promising method to track the assets, although cost and system integration barriers still limit their implementation. In the case of pathology laboratories, introducing at least one barcode reader per workstation is recommended. Tracking an individual specimen with the combined use of printers and code readers accelerates the work at the microtome stations, helping histotechnologists track each block and slide, ensuring the adequate identification and concordance between the individual block and slide labels.
An LIS typically offers laboratories some capability to customize the format and content of their slide labels. As will be further explained in the subsequent sections of the document, the employment of unequivocal 2D barcodes can have a multitude of applications in the proposed digital workflow, significantly reducing the operations time and error rates. The impact of such implementation can be noted starting from the accessioning phases, where the specimen is assigned its unequivocal code that will be used later during the processing and reporting steps. This can further help in the creation of tissue cassettes, in the production of tissue glass slides, in the automatic request of additional histochemical and immunohistochemical (IHC) stains, as well as in the double check that should be carried out at every checkpoint to ensure correspondence among received material, grossed specimen, embedded sample, and cut sections. This is facilitated by the additional use of barcode readers and by the implementation of newly introduced instruments to capture the cut surface directly from the paraffin block, which is at this point essential to guarantee a sustainable and reliable quality control process (see the next section).
Quality control program and definition of checkpoints
Quality control of products from a pathology laboratory is essential to guarantee that a patient receives a correct diagnosis. In Europe, the certification and accreditation of laboratories are not equally and uniformly performed across the territory. Instead, many laboratories design their own quality control program, more or less simplified, often involving only segments of specimen processing adequate to their intent. Although adopting a quality management system is not strictly required in all countries as a prerequisite for implementing DP workflow, laboratories with a robust system of quality management may find the DP workflow easier to implement as they are already aware of the critical control checkpoints through the analog workflow. To support those laboratories that are not yet familiar with quality control programs, a detailed description of some suggested checkpoints suitable for adaptation to each laboratory are provided. The checkpoints described here derive from the need to control the performance of a new instrument in the pipeline—the scanner. They also originate from introducing new standard operative procedures (SOPs), tools, instruments, and quality control of the processes (Figure 1).
Simultaneously, per each workstation, some technical modifications are discussed to facilitate the scanning process and increase the quality of the WSI. We highlight that time loss within the laboratory is frequently motivated by a mismatch of specimens and poor specimen quality (either due to a pre-analytical or analytical factor). Investing in a workflow with a good quality of specimens that are easy to track decreases the time lost, considering that this loss is very difficult to estimate because it is not generally recorded.
During the accessioning phase, specimens that arrive at the pathology laboratory are registered in the LIS and given a case number. Analyzing the classical analog accessioning procedures allows for a critical evaluation of the potential issues that can and do happen. Mistakes can occur in the compilation of paper requests from the submitting department or outside hospital (for internal and external cases, respectively). The staff responsible for the accessioning phase can miss discrepancies between the specimen or slide and the request, or even mismatch this pair. The manual loading of the specimen and associated patient data into the LIS can impair the link with the patient profile present in the hospital information system (HIS), generally due to an inappropriate transcription of a patient’s identification data, eventually causing a duplication of the patients’ profile, and consuming time. In a laboratory with a DP workflow, laboratory personnel have the possibility of completing these accessioning tasks automatically to minimize the risk of errors. The different identification codes (IDs) used in the various subsequent steps play a crucial role here. Similar to a pyramid or hierarchy, different types of IDs are attributed to the patient, and everything associated with this accessioning event, as follows:
- Patient ID
- C ID
- Specimen container ID (entry lab)
- Sample IDs
- Block IDs
- Slide IDs
For DP, a mechanism is needed to get required case information to the administrative department and the pathologist, including the patient ID and demographic information, a description of the specimen or sample, the clinical history, and any questions or requests for the pathologist. In the DP laboratory, the accessioning process is modified and may include some or all of the following (Table 1):
- Specimen/slides arrive in the pathology laboratory with a label containing a code (entry lab, preferentially 2D type) associated with patient and case data.
- By scanning the code on the label of the case, the administrator is able to open the digital request in the pathology LIS automatically, allowing the automatic synchronization of the information from the hospital system or creating the specific page for cases/patients coming from outside.
- A case ID for the specimen is generated.
- The case ID is then used in all sorts of sampled assets generated for that case (cassettes, new slides, special stains, digital slides).
- The administrator can take pictures of both the container and the specimen, and these photos will be attached to the case file.
- All of the documents received together with the specimen are scanned and attached to the case file or directly transmitted to the LIS digitally (e.g., via optical character recognition [OCR]).
|Table 1. Suggested checkpoints at the accessioning workstation|
|1. Specimens/slides are accessioned, using an order entry system, after the scanning of a code that identifies the patient/case, imports all the necessary information from the integrated HIS and opens the digital request: a procedure that introduces automation and consequent reduction of transcription errors.|
|2. A number and the respective identification code are generated for each sample, and the identification code is used for various assets generated for that case: a procedure that allows the tracking of the sample while it is circulating in the laboratory.|
|3. Dedicated personnel take a picture of the container and of the specimen, and those photos are attached to the case file: a procedure that documents the product entering the laboratory as well as the respective identification; this may represent an important medico-legal registry.|
|4. The documents that may be received with the specimen are scanned and attached to the case file or directly transmitted digitally (OCR): a procedure that facilitates access to relevant information that is prevented from being lost in a workstation.|
After accessioning, cases are ready to be macroscopically analyzed, described, and grossed by the pathologist or trained technical staff. As this happens in accessioning, the grossing workstation may be a source of human errors. These errors may include wrong assignment of the macroscopic description and grossing of one patient in the paperwork of another patient; loss of manual transcription of specimen descriptions; deterioration of the numbers on the cassettes; and incongruences among the specimen received, grossed, and subsequently processed in the absence of step-by-step picture documentation.
As in the other workstations, through automation, the DP workflow can help reduce to a minimum the human interference needed in the grossing phase. The previously described process would be as follows (Table 2):
- The grossing operator (e.g., pathologist/resident/pathologist’s assistant) can access the case/patient file by directly scanning the code on the specimen container.
- Pictures are taken of the specimen before it is described and grossed, as well as during grossing, and finally of all tissue cassettes with slices; those images are directly linked to the case using the software integration paths between the LIS and the image capture instrument.
- The grossing operator performs a macroscopic description of the specimen through automated speech recognition systems that report the text in the appropriate section of the case/patient file using the software integrations paths between the LIS and the dictation system instrument.
- The operator can produce cassettes by using a specific printer (preferably laser printer) to assign an identification code corresponding to the particular case, as established during the accessioning, and using the software integration paths between the LIS and the printer. The cassettes and marker media should be appropriately tested to demonstrate the indelibility or impossibility of washing away or removing the identification code. The suggested code is 2D (e.g., QR code), which can include a greater character count (higher data density), require a smaller footprint, and have fewer scan and printer failures than 1D codes.
- An image of the cassette with the grossed specimen should be obtained at the bench, allowing retrieval of this at the following steps.
|Table 2. Suggested checkpoints at the grossing workstation|
|1. Scanning the identification code on the specimen container allows for automatic access to the patient/case data, preventing transcription errors.|
|2. The photographic documentation of the specimen as it is in the container, during grossing, and within the cassettes (for comparison to what is arrived at the embedding station) guarantees the preservation of the case features and identification. The automatic introduction of the photographs into the patient/case file at the LIS prevents mismatches and time loss.|
|3. The macroscopic description of the specimen is dictated and converted to text through voice recognition functions of the LIS or of an instrument connected to the LIS preventing transcription errors and time loss.|
|4. Cassettes are printed with the identification code of the sample to be tracked in subsequent workstations.|
|5. The material inserted in the cassette during grossing can be captured to obtain retrievable pictures at the following steps.|
Grossing-to-processing and processing checkpoints
After the grossing phase, cassettes containing the specimens are ready to be processed. At this stage, further checkpoints may be needed to verify that all the cassettes generated at the grossing workstation are present in the rack to be processed. This double-check is still routinely and primarily done manually in most pathology laboratories. In a DP workflow, this task may be carried out by scanning the codes printed in the cassettes of the rack before they are processed, and checking if all the produced cassettes are submitted to the subsequent phase, integrating the information in the LIS. During the processing phase, both the instruments and programs used should be preferentially recorded through the employment of an appropriately integrated LIS, allowing for tracking the specimens/cases at this workstation. This system can be further deployed to track the usage of reagents for processing, helping in the safe disposal of these reagents. Moreover, the integration with the LIS can further help aggregate specimens and cassettes in different racks based on their processing time and scanner time/protocol requirements (e.g., fast vs. standard processing), or even separate specimens processed in different instruments.
Once the processed specimen inside the cassettes have arrived at the embedding room, operators (technicians) should be able to access the pictures taken during the grossing phase by simply scanning the barcode (Table 2). This will allow them to compare them with the content of cassettes after processing, checking their correspondence to rule out the loss of biological material. Correct embedding may prevent the creation of poor-quality virtual slides. One of the possible issues during the scanning phase is represented by the presence of large fragments, which are more prone to be hydrated during the processing steps and thus more complicated to be captured by the scanner. To address this problem, the fragments should be reduced during the grossing phase, and the embedding checkpoint is essential to control this point. Similarly, tissue fragments well oriented, levelled, and close to each other in paraffin, may constitute good substrates for better-quality glass slides. If the extracted sample to be analyzed is too large to be fitted in a regular glass slide, the recent introduction of dedicated scanners for “macro” glass slides provides the possibility of this solution directly from the grossing room.
The sectioning workstation is a time-consuming phase of the laboratory flow where errors are frequent. Here, automation can facilitate the technician’s work, bringing increased control and fewer errors, and resulting in less time spent. The sectioning workstation is complex and requires the rapid manipulation of specimens and instruments in a consecutive way. The introduction of a slide printer, a code reader, a desktop interface, and similar devices can be initially perceived as a further complication of this step.
Checkpoints can be installed at this workstation depending on the laboratory’s needs and may prevent important errors (Table 3). Moreover, the employment of a slide printer (e.g., laser) at the sectioning station connected to the tracking system should be preferred over “classic” handwritten or printed labels in order to reduce the risk of mismatches. The LIS should also be the source of all the information regarding the types of stains to be performed (i.e., IHC or “special” tinctorial stains) from a specific block. Moreover, the introduction of dedicated instruments to capture the cut surface of each paraffin block could represent an additional checkpoint, helping to further reduce tissue inconsistencies among the blocks and the final glass/virtual slides. The sectioning process should follow the highest operative standards to minimize errors and poor quality in the subsequent scanning phase. Indeed, the irregular thickness of a tissue section, and the presence of holes or scratches and debris erroneously collected from the bath, can impair the correct scan of the final glass slide product. The same is true for sections located at the edges of the slides, which may pass undetected by the scanner. Thus, the sections should be thin enough during the cutting phase and preferentially located in the middle of the physical glass slides to ensure the most appropriate scanning quality. Automatic microtomes may contribute to decreased tissue thickness variations.
|Table 3. Suggested checkpoints at the sectioning workstation|
|1. The code printed on the paraffin block may be scanned to open the case file through the integrated LIS, preventing transcription errors.|
|2. The technician can check how many and which kinds of slides are needed for each block directly on the LIS.|
|3. For each paraffin block, one or more printed glass slides are then generated through a dedicated printer, with all the slides having a unique identifier.|
|4. After sectioning, each paraffin block may be photographed to assess whether all the material emerged on the glass slide/WSI.|
|5. The sectioning phase should follow high operative standards, reducing the risk of artifacts that can impair the scanning phase.|
Staining and mounting checkpoints
Once in the staining workstation, the slides produced in the digital workflow are identified through their code to define which staining protocol they should follow, as per internal LIS prerecorded indications, using automated staining platforms. As in the previous step, the staining process should follow the highest qualitative standards to reduce possible modifications that can interfere with the scanning phase (faint or darker staining, debris/precipitates). For this purpose, implementing an internal checkpoint with daily controls and/or external quality control can help assess the quality of stained slides. Automating the staining may contribute to a stable result, allowing the design of a scanning protocol applicable to most of the slides, avoiding restaining and rescanning slides. The production of consistent staining with a clean background is relevant because it decreases the size of the produced digital slides. A final word is needed to address the mounting process and respective automation. To minimize the interference of the mounting medium in the scanning process, the laboratory must select the mounter, the coverslip type, and respective mounting medium to be used in all sorts of glass slides so that the scanner can be calibrated accordingly. Before the scanning phase, it is of paramount importance to check whether the slide is in an adequate state for scanning. After the staining/cover-slipping phase, it should be dry to prevent scanning problems (e.g., stitching, blurring, out of focus areas). Moreover, the scanning phase can be either affected by the different positions of the coverslips, leading to a misalignment of the slides in the rack. Differences in the type of coverslip can be responsible for a high rate of WSIs being out of focus. The use of automatic mounters obviates variations in the quality of the mounting and prevents errors if an adequate revision of the mounter is provided.
Correcting assigning of the WSI to the case checkpoints
This topic is discussed further in the next section on scanning.
After the sectioning and scanning phases, blocks and slides can be appropriately archived to be retrieved whenever is necessary. This task has been historically performed manually by operators (technicians or laboratory assistants), leading to loss and misplacement of blocks/slides, with obvious medico-legal consequences. Moreover, the wide practice of consulting archival material by all the laboratory workers, including residents and students for didactic purposes, can further complicate the correct positioning of these specimens. Based on these observations, full integration with the LIS and the presence of unique identifiers, both on the blocks and glass slides, allow for a more procedural yet automated archiving of all the biological material, as well as its safe and unbiased retrieval if needed when, for example, a request of external consultation arrives. (For archiving of the WSI, please refer to the data retention policy, discussed later in the section on "Validation of WSI for clinical use.")
Scanner for slide digitization
This section contains some considerations and recommendations for selecting and managing the most appropriate digital slide scanner, given a set or requirements for the scanning phase (Table 4). As in other medical specialties, which have been dramatically changed by the introduction of a wide variety of digital devices for routine daily work, it is not the focus of the present recommendations to draw a meticulous review of the technical characteristics of a scanner, since several studies have already been published on this subject.
|Table 4. Recommendations for the scanning phase|
|1. At any given time, two scanners digitize twice the number of slides compared to a single scanner, and three scanners triple this (e.g., for a caseload of 300 slides per day, employing three scanners with 100-slide capacity could be better than using a single scanner with a 300-slide capacity).|
|2. It is advisable to scan during the daytime, with the lab personnel present to solve unexpected problems.|
|3. Scanning sessions during the night might be problematic in some already established workflows; thus, if there are problems with the scanning process, it might be better to avoid scanning after working hours.|
|4. A single-scanner approach is not recommended when contemplating a daily routine diagnostic workflow.|
|5. Consider the possibility of a continuous loading and eventual prioritization of a batch of slides.|
The most appropriate scanner should be selected based on the needs of the specific laboratory (e.g., primary diagnosis, consultation, education, and research). The following section focuses on the possible impact of such a choice on DP workflow implementation. According to the LEAN approach, as discussed earlier, the positioning of the scanners should follow the logic of an automated workflow and thus be placed as close as possible to the staining and cover-slipping stations, making their implementation in the entire process easier and smoother.
The transition to digital pathology also includes choosing the most appropriate types and numbers of scanners for the lab. Although it is highly dependent on the needs of each specific laboratory, one way to estimate the number of scanners is to review previous DP experiences. In this context, each department should be aware of the expected application of the scanners, the total time required for the scanning process, and the time that can be dedicated to this part of the workflow. Some formulas to calculate the numbers of scanners needed in the lab have been proposed.
However, many variables must be considered when calculating the number of scanners required to digitize the entire slide volume within the same workday, thus not interfering with the turn-around time (TAT). These variables are limited by the technical specifications of the scanners and their related informatics networks (including bandwidth and switches) and the type and location of storage, together with the existing workflow within the lab (i.e., availability of personnel at all times of the day). One of the possible pitfalls in calculating the actual scanning time per slide/batch, and thus the number of scanners required per lab, could be represented by the reported scanning times by each vendor, generally calculated on a specimen tissue of 1.5 cm × 1.5 cm in size and with a local storage solution. However, this is far away from the routine practice of an anatomic pathology laboratory that must accommodate very small pieces of tissue (e.g., biopsies) as well as large surgical specimens, and that may even have the possibility of storing the WSI remotely or in the cloud. Moreover, since the implementation of scanners should not impact the existing workflow and eventually lead to its improvement, there is a need to evaluate a continuous loading capability to preserve the same or similar workloads over time compared to the conventional analog counterpart. This should be coupled with the possibility of prioritizing a specific batch of slides.
A few additional comments are needed at this point. Broadly speaking, scanning during working hours should be preferred for practical and logistical reasons. If there are problems with the scanning process, it might be better to avoid scanning after working hours. For example, it has been reported that the mean scanning time in a routine environment is about six minutes for scanning a slide at an equivalent of 40 × magnification. Therefore, it takes about four hours for one rack of 40 slides, and up to 40 hours to digitize all of the slides that fit inside the scanner (using an AT2, Leica Biosystems, Nussloch, Germany). However, it is well known that the scanning process may stop for several reasons, including sticky glass, connection problems, or software and hardware problems. The result may be an incomplete digitization of slides, with a consequent interruption of routine workflow in the subsequent morning. Based on the previous observations, and since the scanning process should be a continuous workflow in the lab, scanning during the day should be preferred to the overnight approach. This could enable lab personnel to react to the possible technical issues mentioned above. This could even lead to modifications in other parts of the pathology workflow to adapt routine specimen processing to the loading schedules required by the scanners.
In this setting, the laboratory can choose to switch from bulk production of slides at the end of the day to a more continuous production of derived samples. Once the number of scanners needed and the required scanning time has been defined, the laboratory should verify whether the number of working operators employed in the department is sufficient to run the instruments for the specified length of time. Otherwise, the calculation of the necessary personnel is required, considering both 1. the scanning process and 2. any virtual slide quality control. The scanning process mainly consists of loading/unloading slides in the scanner and taking snapshots to ensure that the instrument captures all the material on the glass. On the other hand, the quality control phase is equally important and may be time-consuming, encompassing all the quality check procedures of the final WSI and related data.
The points discussed above pertain to “regular” scanners for bright field microscopy. Other “special” scanners exist, e.g., those for dark field microscopy (immunofluorescence and fluorescence in-situ hybridization [FISH]), as well as those for whole mount slides (macro slides). Because of their highly specific fields of application, they are not the object of these recommendations.
Validation of WSI for clinical use
Several validation studies for WSI have been published, and most show broad concordance between the conventional microscope and the digital diagnosis. Regarding staining and sample types applicable to WSI-based diagnosis, most basic tissue slides stained with hematoxylin and eosin (H&E), as well as most special stains and IHC stains, are expected to be usable. However, they require appropriate validation studies, followed by trial periods until the users have reached an adequate learning level. The recommended validation period for the clinical use of WSI should allow each pathologist to follow a training phase with parallel access to glass and digital slides for each case, with different wash-out intervals of time proposed.
This path for the implementation of WSI for primary diagnosis has been followed by different laboratories worldwide. Some pathologists’ professional societies (e.g., College of American Pathologists) have proposed detailed guidelines for this validation process. These have recently been updated, although they are mainly centered on validating WSIs in the diagnostic setting, not considering all the pre-analytical phases of the digital workflow. Here we discuss further critical points that have recently emerged as impactful in the implementation and validation of WSI. They include the most appropriate visualization devices, assessment of scan quality, tissue coverage of the block, glass slide and virtual slide, and the proper assignment of the WSI to the case and/or the patient.
The visualization chain - The most appropriate monitor and display: The pathologist workstation
The typical pathologist workstation is composed of one computer and two monitors. One monitor displays the LIS, showing patient data and different dashboards, with the possibility to access the patient’s documents or slides. The other monitor is dedicated to the visualization of the WSIs or other images. Several documents have already described all of the features needed to implement the visualization instruments in DP, namely, monitor quality, brightness and contrast, color depth, fidelity, and profiles. Many pathology departments already operate with workstations equipped with high-contrast (e.g., minimum contrast ratio of 1000:1), high–resolution (e.g., 16:10, 27′ diagonal matrix, 2560 × 1600), and bright displays (e.g., a maximum brightness of 300 cd/m2). Some FDA-approved built-in solutions for digital pathology employ medical-grade displays. However, the minimum requirements for DP monitors are still debated, and there is no consensus on how to assess their quality. External sources of variability further complicate this matter, such as the distance from the monitor and the illumination conditions of the room, which makes unbiased comparisons among the different devices available more difficult. This heterogeneity, and the large variety of supply in the digital market, has been recently reviewed, stressing the need for appropriate information of pathologists on this topic due to the complexity of the available technologies, which are changing at a fast pace. Alternatively, an easy-to-access and point-of-use quality assessment tool has been proposed, which tests color accuracy and may be a valuable indicator of the suitability of a particular screen for digital pathology diagnostics. Further validation is needed for its definitive employment in this setting.
Many comments could be made to identify the minimum computer technical requirement at the pathologist’s workstation. Dedicated random access memory (RAM) allows pathologists to visualize the WSI correctly. However, there is no standard in selecting CPU or RAM, as the management of WSIs may be affected by several parameters (e.g., network connection, switches, etc.). A recent mid-range gaming computer will undoubtedly exceed the technical requirements to approach DP.
Scan quality assessment
After digitization, the produced WSI should be checked to ensure appropriate image quality to avoid any technical interference with the final diagnosis. Although most of the routine slides (about 90%), if properly processed, should not present scanning problems, some special slides (e.g., IHC or FISH) would benefit from dedicated scanning protocols and could be affected by more digitization issues. On the other hand, a minority of the “routine” slides (about 10%) could still be affected by scanning issues, stressing the need to adopt alternative protocols to obtain WSI from these challenging specimens. This is based mainly on the assessment of focus quality, which can be partly assisted by the automated metric implemented in some available scan systems but should be performed on every slide to decide whether to rescan the derived sample. This can be done systematically by the lab personnel (e.g., technicians) for every scanning set before assigning the case to a pathologist. Alternatively, the pathologist can perform this check once after case review, requesting a rescan of a glass slide similar to the way that an additional recut or a special stain is ordered in the LIS. However, under real-life conditions, the entire manual check could be rather time-consuming and troublesome, especially in light of the need for subsequent deployment of image analysis algorithms. For this reason, automation of this phase is highly recommended. It can further speed up the digital transition process, ensuring an adequate quality of the scanned slide for potential subsequent AI analyses. Suppose pathologists review a slide with blurry areas. In that case, it is up to their judgment to decide whether these artefacts will interfere with their safe diagnosis of the image, and order rescans as necessary. However, even in this case, it can be challenging for the human naked eye to unmask potential slight imperfections of the scanned slides that can impair the employment of AI algorithms. Even in these cases, the introduction of focus quality assessment and quality control computational tools have been developed. In this setting, however, further validations are needed to implement such algorithms in routine practice.
A critical assumption with using WSI in clinical settings is that scanned slides are completely accurate digital representations of glass slides. Therefore, it is of paramount importance that all tissue fragments present on glass slides be recognized and captured for review on the resulting digital slides.
Typically, two main different images are generated during the digitization process:
- the overview (rendering the macro/slide label files), which is a low-resolution snapshot of the entire glass slide; and
- the “digital image” of the glass slide generated by a microscope camera (rendering baseline tiled image, thumbnail and multiple intermediate tiled images stacked in a pyramid), often acquired at the chosen magnification.
Many WSI devices include systems to detect tissue samples on a slide to limit scanning to relevant tissue-containing areas. For example, some scanners are programmed to omit the blank areas on the slide during the scanning process, where tissue is presumed to be absent, to speed up the process and generate files of smaller size. However, sometimes the tissue detection mechanism can fail to identify small or pale pieces of tissue automatically (e.g., label them as blank areas), or the user may not appropriately select the region containing the entire tissue analyzable. This may lead to potential errors that can cause serious discrepancies between the tissue present on the glass slide and the WSI. In this setting, overview images can help avoid such errors and represent a valuable tool for quality assessment purposes. The macro image provides a low-magnification overview of all the tissue pieces and empty space on the glass slide. It serves mainly to guide the scanner’s tissue detection system, focus-point selection, and subsequent high-resolution digitization of tissue recognized and/or manually selected by an operator. However, the macro image is not necessarily displayed by default by all WSI vendors. Pathologists and laboratory personnel should be adequately trained on how to find and use the macro image as part of essential quality control. An alternative way to ensure that all the available tissue is present on the WSI is to compare the obtained digital slide with the original glass slide. However, this process could be time-consuming and represent a continuing additional workload for the lab personnel (e.g., histotechnologists), avoidable with the proposed double-check practice, with the macro images made by each pathologist when the virtual cases are assigned.
Other possible sources of tissue coverage errors can result from inappropriate placement of tissue sections on the glass slide (e.g., below the label or on the external frame of the slide), which can fall outside the area recognized by the scanner. To address this issue, it is also essential to check the technical specifications of the scanner, specifically regarding the predefined detectable area, which could be insufficient to cover the entirety of the physical glass slide containing tissue. These issues can be addressed by following the point-by-point indications reported above on the sectioning and cover-slipping checkpoints. Finally, very few cases with concordant WSI and macro images can still hide discrepancy issues with the entire amount of tissue sent for the analysis in the lab. This could be addressed by the manual, analog comparison of the tissue block with the macro images and the WSI, as suggested in the earlier section about checkpoints. However, the use of appropriate instruments to take a picture of the cut surface of the block can represent a valuable digital cost-effective alternative to reduce the error rate further. The pathologist can then readily check the three-way concordance among digitized cut surface of the block, macro images, and WSI for every assigned case, reducing the error rate close to zero.
Assignment of images to the correct case/patient file
At the end of the scanning process, one of the most important steps is correctly assigning the digitized slides to the appropriate case/patient. As already mentioned, during the scanning phase a macro image is generated representing a snapshot of the entire glass slide, which usually includes the slide label with identifiers (e.g., case accession number, barcode, text showing a patient name, and slide level or stain details). As per other steps mentioned earlier, a 2D barcode is crucial here as well to allow the scanning system, adequately integrated with the LIS, to link the scanned slides to specific specimens and patients. In some cases, according to institute policies, dedicated personnel can perform a double-check after the automatic assignment of slides. It has been reported that errors in recognizing the printed barcode on the slide (or barcodes printed in the label) may occur, thus preventing the WSI to be matched with the corresponding case. This is due mainly to the poor quality of the printed barcode (or because it is missing a part of the barcode). This ultimately results in a case that is not ready to be reported, with a consequent delay in diagnosis. A checkpoint should be performed at this step to verify that all the stained slides have been digitized and assigned to the correct cases. Usually, scanner vendors create specific folders of “unassigned” (or barcode-less) slides for this purpose.
Topics not fully addressed as of yet
Data retention policies and image storage solutions
Data retention policies and image storage solutions for WSIs still represent open and debated topics, and no strict recommendations have been provided yet, although some suggestions and indications can be found in different documents and guidelines. General recommendations, including those from the College of American Pathologists (CAP), currently advise retaining glass slides for at least 10 years, with suggested periods of retaining digital pathology storage for a period ranging from a few years up to several years. However, regulation on the retention of virtual slides, when used for primary diagnosis, is still lacking. Some documents suggest applying the same indications used for glass slides for WSIs, too. Other recommendations are to keep the digital image for a period of two laboratory inspection cycles in case of any need to review it (e.g., for audits, quality control, medico-legal reasons). However strict and precise European guidelines are still needed to define the minimum period of time for WSI storage. Until then, pathology departments should determine an appropriate retention policy for their digital images.
Closely related to retention policy is the appropriate storage of images; where, how, and which slides must be stored is still an unresolved issue with multiple possibilities that can be adapted to each laboratory’s needs. Regarding the solutions available on the actual storage of the WSIs, organized and redundant storage solutions (e.g., network attached storage [NAS] or redundant array of independent disks [RAID]) are preferred to simple external/internal hard drives, considered negligent by some authors. The possibility of multiple backup copies and disaster recovery procedures should also be kept in consideration. Moreover, the issue related to the file extension of the WSIs that should be employed during the backup is still unsolved. One of the essential requirements is represented by the capability of ensuring the changelessness, the guaranteed future and future-proofing of the data (e.g., the unity of patient or case data and the actual image content), as well as the easy accessibility of the WSIs, even after years. This can be obtained by the use of a DICOM-capable archive, although the eventual loss of quality related to the compression/conversion from one image extension to another is still a matter of debate. Finally, identifying the amount and type of storage needed is important, as it is one of the highest costs when implementing DP and needs to be adapted to the calculated yearly needs of each laboratory.
It has to be emphasized that digital pathology storage needs to be built both interoperable and useful. This requires high-quality datasets, seamless communication across IT systems, and standard data formats. Interoperability is of paramount importance for achieving the full potential of digitization in healthcare and medicine to avoid the risk of having data difficult to exchange, process, and interpret. Interoperability should be technical, syntactic, semantic, and organizational. Regarding this, it has been suggested that enterprise vendor-neutral archiving, composed of hardware and software, could be used to accumulate images directly from various image acquisition sources, with the possibility to manage images and other end-user applications such as EHR, LIS, and other health-related information systems and databases.
Evaluation of the results obtained with the digital transition
The process of DP implementation has designs and consequences that are distinct in each pathology laboratory. Monitoring the effects of the digital transformation of the laboratory is good practice. It may include an analysis of the parameters such as quality control results, turn-around time, workload for technicians and pathologists, ergonomics of each workstation, and the general satisfaction of the staff before and after the implementation. For validation and quality control of the implemented digital workflow, we direct the reader to the recommendations already in use.
Preparing for the subsequent steps after implementing the digital workflow
Having a digital repository that contains clinical and histological information can lay the foundation for using numerous computational pathology tools. The application of image analysis algorithms can allow the identification of specific cell or tissue compartments (e.g., nuclei, mitosis, glands, stroma, among others) for quantification (e.g., cell or mitosis counting) as well as for classification purposes (e.g., grading). Some practical applications of these tools range from helping in the more rigorous scoring of some IHC-stained sections (e.g., programmed death-ligand 1 [PD-L1] and human epidermal growth factor receptor 2 (HER2) scoring), to the quantification of the proliferation index (Ki-67) in many neoplasms (e.g., breast, lymphomas, and neuroendocrine tumours). Moreover, the recent introduction of more sophisticated elaboration algorithms allows further information, starting from the digitized images and integrating the clinical, laboratory, and radiological data to obtain diagnostic, classification, and prognostic hints through the application of the so-called AI. Further simplifications in the work of pathologists are possible in the future, such as the automation of time-consuming repetitive tasks and the extraction of more data from the tissue to support precision medicine.
The present recommendations represent a European guidance for transitioning from a classic, “analog” to a completely digital workflow in every anatomic pathology department. Ten basic principles (Table 5) resulted from the discussion among international experts after implementing the available updated national guidelines. Based on the present document, the anatomic pathology societies of every European country should be able to direct the departments towards DP transition. Updates in the future will provide dedicated indications on the adoption of computer-aided diagnosis and AI tools.
|Table 5. Summary of the recommendations for the implementation of a digital pathology workflow|
|Principle||Type of action|
|1. The transformation of a laboratory toward digital pathology requires a multidisciplinary approach (pathologists, technicians, IT).||Recommendation|
|2. Involve all the team in the transition process toward digital pathology (i.e., during the educational phase).||Recommendation|
|3. Spare valuable resources (e.g., spaces, time, and people) employing the LEAN approach to optimize the process.||Suggestion|
|4. Analyze the potentialities of an LIS or LIMS and be aware of the information resources.||Recommendation|
|5. Start automation of all the possible processes, implementing a tracking system and defining the appropriate checkpoints for every phase.||Recommendation|
|6. Design a quality control program, mapping the necessary quality control steps.||Recommendation|
|7. Choose an appropriate scanner.||Suggestion|
|8. Validate WSI for clinical use.||Recommendation|
|9. Evaluate the impact and results of the digital transformation and other members of the team to perform the same analysis.||Recommendation|
|10. Prepare the next steps for digital pathology implementation after the workflow is well established.||Recommendation|
We would like to thank Umberto Malapelle and Pasquale Pisapia from the University of Naples Federico II for their support in the creation of the iconography of this manuscript.
Conceptualization, F.F. and C.E.; methodology, V.L.; writing—original draft preparation, F.F. and V.L.; writing—review and editing, D.A., R.C., S.L., T.-R.K., M.S., D.R., V.D.M., A.P., N.Z.; supervision, F.F. and C.E. All authors have read and agreed to the published version of the manuscript.
This research received no external funding.
All the data used are present with the manuscript.
Conflicts of interest
Filippo Fraggetta holds a US Patent App. 16/688,613, 2020 Sample imaging and imagery archiving for imagery comparison. David Ameisen is a partner at imginIT, SAS.
- Evans, Andrew J.; Bauer, Thomas W.; Bui, Marilyn M.; Cornish, Toby C.; Duncan, Helena; Glassy, Eric F.; Hipp, Jason; McGee, Robert S. et al. (1 November 2018). "US Food and Drug Administration Approval of Whole Slide Imaging for Primary Diagnosis: A Key Milestone Is Reached and New Questions Are Raised" (in en). Archives of Pathology & Laboratory Medicine 142 (11): 1383–1387. doi:10.5858/arpa.2017-0496-CP. ISSN 0003-9985. http://meridian.allenpress.com/aplm/article/142/11/1383/102979/US-Food-and-Drug-Administration-Approval-of-Whole.
- Snead, David R. J.; Tsang, Yee-Wah; Meskiri, Aisha; Kimani, Peter K.; Crossman, Richard; Rajpoot, Nasir M.; Blessing, Elaine; Chen, Klaus et al. (2016). "Validation of digital pathology imaging for primary histopathological diagnosis" (in en). Histopathology 68 (7): 1063–1072. doi:10.1111/his.12879. ISSN 1365-2559. https://onlinelibrary.wiley.com/doi/abs/10.1111/his.12879.
- Goacher, Edward; Randell, Rebecca; Williams, Bethany; Treanor, Darren (1 January 2017). "The Diagnostic Concordance of Whole Slide Imaging and Light Microscopy: A Systematic Review" (in en). Archives of Pathology & Laboratory Medicine 141 (1): 151–161. doi:10.5858/arpa.2016-0025-RA. ISSN 0003-9985. http://meridian.allenpress.com/aplm/article/141/1/151/65950/The-Diagnostic-Concordance-of-Whole-Slide-Imaging.
- Mukhopadhyay, Sanjay; Feldman, Michael D.; Abels, Esther; Ashfaq, Raheela; Beltaifa, Senda; Cacciabeve, Nicolas G.; Cathro, Helen P.; Cheng, Liang et al. (1 January 2018). "Whole Slide Imaging Versus Microscopy for Primary Diagnosis in Surgical Pathology: A Multicenter Blinded Randomized Noninferiority Study of 1992 Cases (Pivotal Study)" (in en). American Journal of Surgical Pathology 42 (1): 39–52. doi:10.1097/PAS.0000000000000948. ISSN 0147-5185. PMC PMC5737464. PMID 28961557. https://journals.lww.com/00000478-201801000-00006.
- Evans, Andrew J.; Salama, Mohamed E.; Henricks, Walter H.; Pantanowitz, Liron (1 July 2017). "Implementation of Whole Slide Imaging for Clinical Purposes: Issues to Consider From the Perspective of Early Adopters" (in en). Archives of Pathology & Laboratory Medicine 141 (7): 944–959. doi:10.5858/arpa.2016-0074-OA. ISSN 0003-9985. http://meridian.allenpress.com/aplm/article/141/7/944/65708/Implementation-of-Whole-Slide-Imaging-for-Clinical.
- Treanor, D.; Williams, B.; Hussain, B. et al. (July 2018). "The Leeds Guide to Digital Pathology" (PDF). University of Leeds. https://www.virtualpathology.leeds.ac.uk/research/clinical/docs/2018/pdfs/18778_Leeds%20Guide%20to%20Digital%20Pathology_Brochure_A4_final_hi.pdf. Retrieved 10 November 2021.
- LabPON Laboratorium Pathologie Oost-Nederland (October 2016). "How to Go Digital in Pathology" (PDF). LabPON Laboratorium Pathologie Oost-Nederland. https://www.usa.philips.com/c-dam/b2bhc/us/landing-pages/pdxus/how-to-go-digital-in-pathology.pdf. Retrieved 10 November 2021.
- Retamero, Juan Antonio; Aneiros-Fernandez, Jose; del Moral, Raimundo G. (1 February 2020). "Complete Digital Pathology for Routine Histopathology Diagnosis in a Multicenter Hospital Network" (in en). Archives of Pathology & Laboratory Medicine 144 (2): 221–228. doi:10.5858/arpa.2018-0541-OA. ISSN 0003-9985. http://meridian.allenpress.com/aplm/article/144/2/221/433661/Complete-Digital-Pathology-for-Routine.
- Fraggetta, Filippo; Garozzo, Salvatore; Zannoni, GianFranco; Pantanowitz, Liron; Rossi, EstherDiana (2017). "Routine digital pathology workflow: The Catania experience" (in en). Journal of Pathology Informatics 8 (1): 51. doi:10.4103/jpi.jpi_58_17. ISSN 2153-3539. PMC PMC5760840. PMID 29416914. http://www.jpathinformatics.org/text.asp?2017/8/1/51/221132.
- Fraggetta, Filippo; Caputo, Alessandro; Guglielmino, Rosa; Pellegrino, Maria Giovanna; Runza, Giampaolo; L'Imperio, Vincenzo (16 October 2021). "A Survival Guide for the Rapid Transition to a Fully Digital Workflow: The “Caltagirone Example”" (in en). Diagnostics 11 (10): 1916. doi:10.3390/diagnostics11101916. ISSN 2075-4418. PMC PMC8534326. PMID 34679614. https://www.mdpi.com/2075-4418/11/10/1916.
- Eloy, Catarina; Vale, João; Curado, Mónica; Polónia, António; Campelos, Sofia; Caramelo, Ana; Sousa, Rui; Sobrinho-Simões, Manuel (15 November 2021). "Digital Pathology Workflow Implementation at IPATIMUP" (in en). Diagnostics 11 (11): 2111. doi:10.3390/diagnostics11112111. ISSN 2075-4418. PMC PMC8620597. PMID 34829458. https://www.mdpi.com/2075-4418/11/11/2111.
- Sinard, John H.; Castellani, William J.; Wilkerson, Myra L.; Henricks, Walter H. (1 March 2015). "Stand-alone Laboratory Information Systems Versus Laboratory Modules Incorporated in the Electronic Health Record" (in en). Archives of Pathology & Laboratory Medicine 139 (3): 311–318. doi:10.5858/arpa.2013-0711-SO. ISSN 1543-2165. http://meridian.allenpress.com/aplm/article/139/3/311/193668/Standalone-Laboratory-Information-Systems-Versus.
- Sepulveda, Jorge L.; Young, Donald S. (1 August 2013). "The Ideal Laboratory Information System" (in en). Archives of Pathology & Laboratory Medicine 137 (8): 1129–1140. doi:10.5858/arpa.2012-0362-RA. ISSN 1543-2165. http://meridian.allenpress.com/aplm/article/137/8/1129/65418/The-Ideal-Laboratory-Information-System.
- Asa, SylviaL; Bodén, AnnaC; Treanor, Darren; Jarkman, Sofia; Lundström, Claes; Pantanowitz, Liron (2019). "2020 vision of digital pathology in action" (in en). Journal of Pathology Informatics 10 (1): 27. doi:10.4103/jpi.jpi_31_19. ISSN 2153-3539. PMC PMC6704755. PMID 31516758. http://www.jpathinformatics.org/text.asp?2019/10/1/27/264400.
- Lujan, Giovanni; Quigley, JenniferC; Hartman, Douglas; Parwani, Anil; Roehmholdt, Brian; Meter, BryanVan; Ardon, Orly; Hanna, MatthewG et al. (2021). "Dissecting the business case for adoption and implementation of digital pathology: A white paper from the digital pathology association" (in en). Journal of Pathology Informatics 12 (1): 17. doi:10.4103/jpi.jpi_67_20. ISSN 2153-3539. PMC PMC8240548. PMID 34221633. http://www.jpathinformatics.org/text.asp?2021/12/1/17/313004.
- Hanna, Matthew G.; Reuter, Victor E.; Ardon, Orly; Kim, David; Sirintrapun, Sahussapont Joseph; Schüffler, Peter J.; Busam, Klaus J.; Sauter, Jennifer L. et al. (1 November 2020). "Validation of a digital pathology system including remote review during the COVID-19 pandemic" (in en). Modern Pathology 33 (11): 2115–2127. doi:10.1038/s41379-020-0601-5. ISSN 0893-3952. PMC PMC7306935. PMID 32572154. http://www.nature.com/articles/s41379-020-0601-5.
- Zarbo, Richard J. (1 September 2012). "Creating and Sustaining a Lean Culture of Continuous Process Improvement" (in en). American Journal of Clinical Pathology 138 (3): 321–326. doi:10.1309/AJCP2QY1XGKTSNQF. ISSN 0002-9173. https://academic.oup.com/ajcp/article-lookup/doi/10.1309/AJCP2QY1XGKTSNQF.
- Tuthill, JMark; Friedman, BruceA; Splitz, Andrew; Balis, UlyssesJ (2014). "The laboratory information system functionality assessment tool: Ensuring optimal software support for your laboratory" (in en). Journal of Pathology Informatics 5 (1): 7. doi:10.4103/2153-3539.127819. ISSN 2153-3539. PMC PMC3986538. PMID 24741466. http://www.jpathinformatics.org/text.asp?2014/5/1/7/127819.
- Petrides, Athena K.; Bixho, Ida; Goonan, Ellen M.; Bates, David W.; Shaykevich, Shimon; Lipsitz, Stuart R.; Landman, Adam B.; Tanasijevic, Milenko J. et al. (1 March 2017). "The Benefits and Challenges of an Interfaced Electronic Health Record and Laboratory Information System: Effects on Laboratory Processes" (in en). Archives of Pathology & Laboratory Medicine 141 (3): 410–417. doi:10.5858/arpa.2016-0146-OA. ISSN 0003-9985. http://meridian.allenpress.com/aplm/article/141/3/410/65705/The-Benefits-and-Challenges-of-an-Interfaced.
- Krupinski, ElizabethA (2010). "Optimizing the pathology workstation "cockpit": Challenges and solutions" (in en). Journal of Pathology Informatics 1 (1): 19. doi:10.4103/2153-3539.70708. ISSN 2153-3539. PMC PMC2956171. PMID 21031008. http://www.jpathinformatics.org/text.asp?2010/1/1/19/70708.
- Dash, RajeshC; Jones, Nicholas; Merrick, Riki; Haroske, Gunter; Harrison, James; Sayers, Craig; Haarselhorst, Nick; Wintell, Mikael et al. (2021). "Integrating the health-care enterprise pathology and laboratory medicine guideline for digital pathology interoperability" (in en). Journal of Pathology Informatics 12 (1): 16. doi:10.4103/jpi.jpi_98_20. ISSN 2153-3539. PMC PMC8240547. PMID 34221632. http://www.jpathinformatics.org/text.asp?2021/12/1/16/311870.
- Roy, Somak; Pfeifer, John D.; LaFramboise, William A.; Pantanowitz, Liron (1 September 2016). "Molecular digital pathology: progress and potential of exchanging molecular data" (in en). Expert Review of Molecular Diagnostics 16 (9): 941–947. doi:10.1080/14737159.2016.1206472. ISSN 1473-7159. https://www.tandfonline.com/doi/full/10.1080/14737159.2016.1206472.
- Phelan, Sine M (2014). "Impact of the introduction of a novel automated embedding system on quality in a university hospital histopathology department" (in en). Journal of Histology and Histopathology 1 (1): 3. doi:10.7243/2055-091X-1-3. ISSN 2055-091X. http://www.hoajonline.com/histology/2055-091X/1/3.
- Hanna, Matthew G.; Pantanowitz, Liron (1 March 2016). "Bar Coding and Tracking in Pathology". Clinics in Laboratory Medicine 36 (1): 13–30. doi:10.1016/j.cll.2015.09.003. ISSN 1557-9832. PMID 26851661. https://pubmed.ncbi.nlm.nih.gov/26851661.
- Bostwick, David G. (1 October 2013). "Radiofrequency identification specimen tracking in anatomical pathology: pilot study of 1067 consecutive prostate biopsies". Annals of Diagnostic Pathology 17 (5): 391–402. doi:10.1016/j.anndiagpath.2013.04.007. ISSN 1532-8198. PMID 23796559. https://pubmed.ncbi.nlm.nih.gov/23796559.
- Lou, Jerry J.; Andrechak, Gary; Riben, Michael; Yong, William H. (2011). "A review of radio frequency identification technology for the anatomic pathology or biorepository laboratory: Much promise, some progress, and more work needed". Journal of Pathology Informatics 2: 34. doi:10.4103/2153-3539.83738. ISSN 2153-3539. PMC 3162748. PMID 21886890. https://pubmed.ncbi.nlm.nih.gov/21886890.
- Snyder, Susan R.; Favoretto, Alessandra M.; Derzon, James H.; Christenson, Robert H.; Kahn, Stephen E.; Shaw, Colleen S.; Baetz, Rich Ann; Mass, Diana et al. (1 September 2012). "Effectiveness of barcoding for reducing patient specimen and laboratory testing identification errors: A Laboratory Medicine Best Practices systematic review and meta-analysis" (in en). Clinical Biochemistry 45 (13-14): 988–998. doi:10.1016/j.clinbiochem.2012.06.019. PMC PMC4518452. PMID 22750145. https://linkinghub.elsevier.com/retrieve/pii/S0009912012003232.
- L'Imperio, Vincenzo; Gibilisco, Fabio; Fraggetta, Filippo (2021). "What is Essential is (No More) Invisible to the Eyes: The Introduction of BlocDoc in the Digital Pathology Workflow". Journal of Pathology Informatics 12: 32. doi:10.4103/jpi.jpi_35_21. ISSN 2229-5089. PMC 8529340. PMID 34760329. https://pubmed.ncbi.nlm.nih.gov/34760329.
- Pantanowitz, Liron; Farahani, Navid; Parwani, Anil (1 June 2015). "Whole slide imaging in pathology: advantages, limitations, and emerging perspectives" (in en). Pathology and Laboratory Medicine International: 23–33. doi:10.2147/PLMI.S59826. ISSN 1179-2698. http://www.dovepress.com/whole-slide-imaging-in-pathology-advantages-limitations-and-emerging-p-peer-reviewed-article-PLMI.
- Janowczyk, Andrew; Zuo, Ren; Gilmore, Hannah; Feldman, Michael; Madabhushi, Anant (1 December 2019). "HistoQC: An Open-Source Quality Control Tool for Digital Pathology Slides" (in en). JCO Clinical Cancer Informatics 3: 1–7. doi:10.1200/CCI.18.00157. ISSN 2473-4276. PMC PMC6552675. PMID 30990737. https://ascopubs.org/doi/10.1200/CCI.18.00157.
- Ferrini, Francesco; Sannino, Gianpaolo; Chiola, Carlo; Capparé, Paolo; Gastaldi, Giorgio; Gherlone, Enrico (14 February 2019). "Influence of Intra-Oral Scanner (I.O.S.) on The Marginal Accuracy of CAD/CAM Single Crowns" (in en). International Journal of Environmental Research and Public Health 16 (4): 544. doi:10.3390/ijerph16040544. ISSN 1660-4601. PMC PMC6406818. PMID 30769768. http://www.mdpi.com/1660-4601/16/4/544.
- Hufnagl, Peter; Zwönitzer, Ralf; Haroske, Gunter (8 September 2018). "Guidelines Digital Pathology for Diagnosis on (and Reports of) Digital Images Version 1.0 Bundesverband deutscher Pathologen e.V. (Federal Association of German Pathologist)" (in en). Diagnostic Pathology 4 (1): 266. doi:10.17629/WWW.DIAGNOSTICPATHOLOGY.EU-2018-4:266. http://www.diagnosticpathology.eu/content/index.php/dpath/article/view/266.
- Cross, S.; Furness, P.; Igali, L. et al. (January 2018). "Best practice recommendations for implementing digital pathology" (PDF). The Royal College of Pathologists. https://www.rcpath.org/uploads/assets/f465d1b3-797b-4297-b7fedc00b4d77e51/Best-practice-recommendations-for-implementing-digital-pathology.pdf. Retrieved 10 November 2021.
- Hanna, Matthew G.; Reuter, Victor E.; Hameed, Meera R.; Tan, Lee K.; Chiang, Sarah; Sigel, Carlie; Hollmann, Travis; Giri, Dilip et al. (1 July 2019). "Whole slide imaging equivalency and efficiency study: experience at a large academic center" (in en). Modern Pathology 32 (7): 916–928. doi:10.1038/s41379-019-0205-0. ISSN 0893-3952. http://www.nature.com/articles/s41379-019-0205-0.
- L’Imperio, Vincenzo; Brambilla, Virginia; Cazzaniga, Giorgio; Ferrario, Franco; Nebuloni, Manuela; Pagni, Fabio (1 June 2021). "Digital pathology for the routine diagnosis of renal diseases: a standard model" (in en). Journal of Nephrology 34 (3): 681–688. doi:10.1007/s40620-020-00805-1. ISSN 1121-8428. PMC PMC8192318. PMID 32683656. https://link.springer.com/10.1007/s40620-020-00805-1.
- Azam, Ayesha S; Miligy, Islam M; Kimani, Peter K-U; Maqbool, Heeba; Hewitt, Katherine; Rajpoot, Nasir M; Snead, David R J (1 July 2021). "Diagnostic concordance and discordance in digital pathology: a systematic review and meta-analysis" (in en). Journal of Clinical Pathology 74 (7): 448–455. doi:10.1136/jclinpath-2020-206764. ISSN 0021-9746. PMC PMC8223673. PMID 32934103. https://jcp.bmj.com/lookup/doi/10.1136/jclinpath-2020-206764.
- Thorstenson, Sten; Molin, Jesper; Lundström, Claes (2014). "Implementation of large-scale routine diagnostics using whole slide imaging in Sweden: Digital pathology experiences 2006-2013" (in en). Journal of Pathology Informatics 5 (1): 14. doi:10.4103/2153-3539.129452. ISSN 2153-3539. PMC PMC4023034. PMID 24843825. http://www.jpathinformatics.org/text.asp?2014/5/1/14/129452.
- Williams, Bethany Jill; Treanor, Darren (1 July 2020). "Practical guide to training and validation for primary diagnosis with digital pathology" (in en). Journal of Clinical Pathology 73 (7): 418–422. doi:10.1136/jclinpath-2019-206319. ISSN 0021-9746. http://jcp.bmj.com/lookup/doi/10.1136/jclinpath-2019-206319.
- Pantanowitz, Liron; Sinard, John H.; Henricks, Walter H.; Fatheree, Lisa A.; Carter, Alexis B.; Contis, Lydia; Beckwith, Bruce A.; Evans, Andrew J. et al. (1 December 2013). "Validating Whole Slide Imaging for Diagnostic Purposes in Pathology: Guideline from the College of American Pathologists Pathology and Laboratory Quality Center" (in en). Archives of Pathology & Laboratory Medicine 137 (12): 1710–1722. doi:10.5858/arpa.2013-0093-CP. ISSN 0003-9985. PMC PMC7240346. PMID 23634907. http://www.archivesofpathology.org/doi/abs/10.5858/arpa.2013-0093-CP.
- Evans, Andrew J.; Brown, Richard W.; Bui, Marilyn M.; Chlipala, Elizabeth A.; Lacchetti, Christina; Milner, Danny A.; Pantanowitz, Liron; Parwani, Anil V. et al. (18 May 2021). "Validating Whole Slide Imaging Systems for Diagnostic Purposes in Pathology: Guideline Update From the College of American Pathologists in Collaboration With the American Society for Clinical Pathology and the Association for Pathology Informatics" (in en). Archives of Pathology & Laboratory Medicine. doi:10.5858/arpa.2020-0723-CP. ISSN 1543-2165. https://meridian.allenpress.com/aplm/article/doi/10.5858/arpa.2020-0723-CP/464968/Validating-Whole-Slide-Imaging-Systems-for.
- Abel, Jacob T.; Ouillette, Peter; Williams, Christopher L.; Blau, John; Cheng, Jerome; Yao, Keluo; Lee, Winston Y.; Cornish, Toby C. et al. (2020). "Display Characteristics and Their Impact on Digital Pathology: A Current Review of Pathologists' Future "Microscope"". Journal of Pathology Informatics 11: 23. doi:10.4103/jpi.jpi_38_20. ISSN 2229-5089. PMC 7518209. PMID 33042602. https://pubmed.ncbi.nlm.nih.gov/33042602.
- Wright, A. (2021). "Point of Use QA Pathology". National Pathology Imaging Cooperative. https://www.virtualpathology.leeds.ac.uk/research/systems/pouqa/pathology/. Retrieved 10 November 2021.
- Kohlberger, Timo; Liu, Yun; Moran, Melissa; Chen, Po-Hsuan Cameron; Brown, Trissia; Hipp, Jason D.; Mermel, Craig H.; Stumpe, Martin C. (2019). "Whole-Slide Image Focus Quality: Automatic Assessment and Impact on AI Cancer Detection". Journal of Pathology Informatics 10: 39. doi:10.4103/jpi.jpi_11_19. ISSN 2229-5089. PMC 6939343. PMID 31921487. https://pubmed.ncbi.nlm.nih.gov/31921487.
- Senaras, Caglar; Niazi, M. Khalid Khan; Lozanski, Gerard; Gurcan, Metin N. (2018). "DeepFocus: Detection of out-of-focus regions in whole slide digital images using deep learning". PloS One 13 (10): e0205387. doi:10.1371/journal.pone.0205387. ISSN 1932-6203. PMC 6201886. PMID 30359393. https://pubmed.ncbi.nlm.nih.gov/30359393.
- Hosseini, Mahdi S.; Brawley-Hayes, Jasper A. Z.; Zhang, Yueyang; Chan, Lyndon; Plataniotis, Konstantinos; Damaskinos, Savvas (1 January 2020). "Focus Quality Assessment of High-Throughput Whole Slide Imaging in Digital Pathology". IEEE Transactions on Medical Imaging 39 (1): 62–74. doi:10.1109/TMI.2019.2919722. ISSN 0278-0062. https://ieeexplore.ieee.org/document/8725582/.
- Fraggetta, Filippo; Yagi, Yukako; Garcia-Rojo, Marcial; Evans, Andrew J.; Tuthill, J. Mark; Baidoshvili, Alexi; Hartman, Douglas J.; Fukuoka, Junya et al. (2018). "The Importance of eSlide Macro Images for Primary Diagnosis with Whole Slide Imaging". Journal of Pathology Informatics 9: 46. doi:10.4103/jpi.jpi_70_18. ISSN 2229-5089. PMC 6319037. PMID 30662792. https://pubmed.ncbi.nlm.nih.gov/30662792.
- Council on Scientific Affairs (September 2020). "Policy PP. Minimum Period of Retention of Laboratory Records and Materials" (PDF). CAP Policy Manual. College of American Pathologists. https://elss.cap.org/elss/ShowProperty?nodePath=/UCMCON/Contribution%20Folders/WebApplications/pdf/retention-laboratory-records-and-materials.pdf. Retrieved 10 November 2021.
- Chlipala, E.; Elin, J.; Eichhorn, O. et al. (n.d.). "Archival and Retrieval in Digital Pathology Systems" (PDF). Digital Pathology Association. https://digitalpathologyassociation.org/_data/cms_files/files/Archival_and_Retrieval_in_Digital_Pathology_Systems.pdf. Retrieved 13 November 2021.
- Stathonikos, Nikolaos; Nguyen, Tri Q.; van Diest, Paul J. (1 July 2021). "Rocky road to digital diagnostics: implementation issues and exhilarating experiences". Journal of Clinical Pathology 74 (7): 415–420. doi:10.1136/jclinpath-2020-206715. ISSN 1472-4146. PMID 32988997. https://pubmed.ncbi.nlm.nih.gov/32988997.
- Lehne, Moritz; Sass, Julian; Essenwanger, Andrea; Schepers, Josef; Thun, Sylvia (2019). "Why digital medicine depends on interoperability". NPJ digital medicine 2: 79. doi:10.1038/s41746-019-0158-1. ISSN 2398-6352. PMC 6702215. PMID 31453374. https://pubmed.ncbi.nlm.nih.gov/31453374.
- Pantanowitz, Liron; Sharma, Ashish; Carter, AlexisB; Kurc, Tahsin; Sussman, Alan; Saltz, Joel (2018). "Twenty years of digital pathology: An overview of the road travelled, what is on the horizon, and the emergence of vendor-neutral archives" (in en). Journal of Pathology Informatics 9 (1): 40. doi:10.4103/jpi.jpi_69_18. ISSN 2153-3539. PMC PMC6289005. PMID 30607307. http://www.jpathinformatics.org/text.asp?2018/9/1/40/245903.
- Janowczyk, Andrew; Madabhushi, Anant (2016). "Deep learning for digital pathology image analysis: A comprehensive tutorial with selected use cases". Journal of Pathology Informatics 7: 29. doi:10.4103/2153-3539.186902. ISSN 2229-5089. PMC 4977982. PMID 27563488. https://pubmed.ncbi.nlm.nih.gov/27563488.
- Aeffner, Famke; Zarella, Mark D.; Buchbinder, Nathan; Bui, Marilyn M.; Goodman, Matthew R.; Hartman, Douglas J.; Lujan, Giovanni M.; Molani, Mariam A. et al. (2019). "Introduction to Digital Image Analysis in Whole-slide Imaging: A White Paper from the Digital Pathology Association". Journal of Pathology Informatics 10: 9. doi:10.4103/jpi.jpi_82_18. ISSN 2229-5089. PMC 6437786. PMID 30984469. https://pubmed.ncbi.nlm.nih.gov/30984469.
- Racoceanu, D.; Capron, F. (1 June 2015). "Towards semantic-driven high-content image analysis: An operational instantiation for mitosis detection in digital histopathology" (in en). Computerized Medical Imaging and Graphics 42: 2–15. doi:10.1016/j.compmedimag.2014.09.004. https://linkinghub.elsevier.com/retrieve/pii/S0895611114001487.
- Cui, Miao; Zhang, David Y. (1 April 2021). "Artificial intelligence and computational pathology" (in en). Laboratory Investigation 101 (4): 412–422. doi:10.1038/s41374-020-00514-0. ISSN 0023-6837. PMC PMC7811340. PMID 33454724. http://www.nature.com/articles/s41374-020-00514-0.
- Racoceanu, Daniel; Capron, Frédérique (26 April 2016). "Semantic Integrative Digital Pathology: Insights into Microsemiological Semantics and Image Analysis Scalability" (in en). Pathobiology 83 (2-3): 148–155. doi:10.1159/000443964. ISSN 1015-2008. https://www.karger.com/Article/FullText/443964.
This presentation is faithful to the original, with only a few minor changes to presentation, spelling, and grammar. In some cases important information was missing from the references, and that information was added.