APW-EDM White Paper

Foreword

This is the collaborative space of the white paper "Anatomic Pathology Workflow in an Era of Digital Medicine (APW-EDM)"

The latest draft of the white paper is downloadable from this folder

Temporary documents from Nicholas Jones available via dropbox

Current contributors are:

• Raj C. Dash, MD,
• Riki Merrick,
• Francesca Frexia,
• Francesca Vanzo, ,
• John D Nolan,
• Dan Rutz,
• Nicholas C. Jones,
• Gunter Haroske,
• François Macary

Introduction

This document, the IHE PaLM “Anatomic Pathology Workflow in an Era of Digital Imaging” White Paper, describes use cases, data elements, actors, and transactions necessary to support anatomic pathology workflows that leverage digital technologies. Abbreviation for the title is APW-EDM.

This white paper lays out a collection of future integration profiles to address how vended systems for supporting anatomic pathology workflows interact and is intended to replace the former Anatomic Pathology Workflow (APW) profile. The first objective is to better take into account the latest advancements in digital imaging, spanning but not necessarily limited to two primary aspects of anatomic pathology workflow:

• the gross/macroscopic examination leveraging state-of-the art more interoperable imaging modalities,
• the histologic/microscopic examination taking advantage of whole slide imaging technology.

The second objective is to break down the APW profile into a set of smaller, easier to implement building blocks, each of these focusing on one key aspect of the anatomic pathology digital workflow. The expectation is to collect sufficient feedback from all stakehoders of digital pathology (vendors, pathologists, institutions) in order to confirm and/or refine this set of profiles before starting to build them as new supplements to the Pathology and Laboratory Medicine Technical Framework.

Open Issues

• APW-EDM-01: Involve industrial offer for archiving and communication of whole slide images, as well as slide scanner manufacturers
• APW-EDM-02: Check the market for imaging modalities, PACS, and image reviewing and annotating software solutions, supporting the DICOM standard with the DICOM 122 supplement “Specimen Module and Revised Pathology SOP Classes” and the DICOM 145 supplement “Whole Slide Microscopic Image IOD and SOP Classes”

Closed Issues

• APW-EDM-03: Tissue microarrays may break the current specimen model. DICOM does cover this as a use case. Tissue microarrays are not used in clinical care, and so are left out of the scope of the profiles that will be derived from this white paper.

Use Cases

Use Case #1: Image Slides for Secondary Review / Consultation

Contributor assigned: Nicholas C. Jones

There are numerous contexts for consultation requests which have subtle, but important differences. Understanding these contexts and the pertinent variations for workflow are necessary to understand systems design, management, and operations for secondary reviews. (Temp Note: See NCJ's Consultation Workshop ppt from 2015 DPA conf for further details.) Terminology note: we will call the requesting party the requester, and the person fulfilling that request the consultant. Also note that all references to "second opinions" could also mean third, or fourth opinions, especially if the original primary/final diagnosis was in disagreement with the true second opinion.

• 1.a. Pathologist requested consults, pre-primary diagnosis.
  • 1.a.1. Intrainstitutional: The pathologist is requesting a second opinion (through digital review) to another pathologist at their institution, either at the same site or another site in their network. In the case where the pathologist was already viewing the slides digitally (primary diagnosis), this would likely be in the same system (i.e. uses synchronous or asynchronous consultation functions within their IMS/PACS). This is likely due to the case being challenging, or otherwise meriting the review of a sub-specialist (or different kind of subspecialist, such as a gyn-specialist requesting the opinion of a dermatopathologist's consultation).
    • Note that this may or may not trigger requests for scans, would not trigger billing for the patient, but may trigger records or documents internal to that organization for tracking consultation between pathologists within the organization.
    • Note that due to the internal nature, it can be assumed the consultant will have access to all electronic documentation (EHR, APLIS, etc.), but that there are likely to be either text based messages and annotations (for asynchronous consultation) and/or verbal communication between requester and consultant in the case of synchronous consultation.
    • Much of the value of this process centers around efficiency (compared to glass slide processes) is in limiting "phone or email tag" issues, maximizing communication ("this ROI is what worries me"), and generally improving organization and documentation of the process.
  • 1.a.2. Interinstitutional: The requester pathologist is requesting consultation from an outside hospital, usually due to the challenge of the case or the need for an external subspecialist.
    • Slides may need to be scanned if the requester was using the microscope. The requester pathologists may only request a subset of the slides to be scanned (i.e. if 9 slides are normal, and 2 are questionable, they may scan and send just the 2 questionable slides to the consultant.) Note that in this context, the scanning processes will likely be considered high priority or rush requests; systems should allow for differences in priority of scan requests.
    • The request to the consultant will be coupled with consultant's summary of the patient's clinical history, notes and questions (i.e. "is this a microfocus of carcinoma?" or "my differential is diagnosis A vs diagnosis B.") along with printed or saved documentation (i.e. preliminary report including items like frozen section or rapid fine needle aspiration interpretations, reports from surgeons or radiology imaging processes, other images from AP processes such as gross images, IF, EM, etc.) and possibly blocks. Inclusion of profiles like APSR here will be highly beneficial, especially if APSR profiles support itemizing both all assets that exist on a case as well as what was sent for consultation.
    • Note that the consulting institution will need to match images to clinical history and documentation. The consulting institution may also have quality control procedures to check image quality, and may need to troubleshoot the process by asking for rescans or other clinical information.
    • There are potentially significant differences in process here based on how image distribution and viewing are intended to function. This may involve uploading WSIs and other data to a thin client host, uploading data to a repository for download to the consultant group, or sending disks (i.e. flash drives or DVDs) with the data.
    • Much of the value proposition in this process revolves around reducing turn-around-time of processes versus mailing glass slides, further coordinating the consultation process (which requires some collaboration between the requesting site and consultant site), and improving on the consultation product to the requester (i.e. requester gets the consultant's report in the end, but can also see annotations or other notes to see in more detail why the consultant interpreted the case in such a way). These value propositions also merit noting that sending physical media can potentially be viewed as "the worst of both worlds" if they require both imaging processes and still have the long shipping time related issues and risks.
• 1.b. Patient requested second opinion, post-primary diagnosis.
  • These contexts can be generally assumed to be going to a different institution. While there may be less pressure on the original institution for timeliness, the patient may be waiting on making a treatment decision based on the second opinion, so process turn-around-time is still very important. But because the patient is likely requesting the review sent to a specific institution or other pathologist, it should not be assumed that there is a functional telepathology pipeline between the two institutions.
  • If there are no specific instructions accompanying the request (other than send this case X to institution/pathologist Y for second opinion), the onus is on the institution sending the material to gather materials (documents, images, scanned slides) for send out. The original hospital may wish to image some or all cases for such send-outs either to archive such materials (to mitigate risk of slides being damaged or lost in transit by having digital copies), or to prevent such risks by sending WSIs instead. There may be less benefit of this process context to the consultant pathologist or institution (other than mitigating risk of damage or loss of glass slides) unless workflows are set up to improve turn-around-time, organization of data, or some other benefit to the consultants. It is important to consider the consultants as stakeholders in these processes as well.
    • Again, exact processes will vary based on image distribution (i.e. whether WSIs and other clinical data are made available for download, sent via physical media such as DVDs, or other methods).
• 1.c. Physician requested second opinion, post-primary diagnosis (NOTE FROM NCJ: Dr. Dash: can you verify that we should separate this out from 1b or should we put it as a sub-category under 1b, since the physician here is going to get pt consent for the consultation anyways? In my experience though, these are of noteworthy difference both in clinical processes and in likelihood of impact /avg value from 1b.)
  • These requests usually come from (non-pathologist) physicians prior to making treatment decisions with the patient. Some physicians (such as oncologists) do this routinely for all new patients, as even a low probability of changing the diagnosis can yield huge value for the patient by preventing unnecessary or inappropriate treatments.
  • It is common for the physician to request the second opinion be sent to the their institution (especially common in academic medical centers, as the clinician may interact with the patient on tumor boards or future biopsies or excisions), or have preferred consultants outside their institution (more common in smaller institutions). Retaining images of prior slides are of significant clinical benefit to pathologists that will review future biopsies or excisions for the patient, as glass slides are generally sent back to the original institution within a specified timeframe. (NOTE FROM NCJ: In the US, the norm is 30 days, though I believe this is a norm rather than a law. Are there differences in practice in Europe and other areas of the world?)
• 1.d. Second-opinion broker requests, post-primary diagnosis.
  • These are a hybridization of 1b and 1c. Patients may request the services of second opinion brokers to organize consultation on all of their case, including pathology but also radiology, oncology, and other areas. The second opinion brokers provide case management services, and may delegate different aspects of the overall consultation to different groups.
  • Differences in the organization of these processes merit their mention here, as the group managing the consultation is external both to the original diagnosing institution and the consultant pathologists. Workflows and data pipelines may either go through the second opinion broker's systems or otherwise need to be coordinated with the second opinion broker organization who act as proxies on behalf of the patient.
• 1.e. Legal cases, post-primary diagnosis.
  • In these cases, imaging may occur to archive the case prior to glass slide send-out for consultation, or to send out the images to legal teams which will seek consultants.
  • Workflows for legal cases are significantly different than clinical processes, but may be less urgent (for clinical care purposes) than other contexts discussed above. (NOTE FROM NCJ: I am not familiar enough with these processes to go into depth here. It's even worth questioning that we want this in the profile, as it's only semi-clinical. But I think there's at least benefit in mentioning that this is something to consider for system and workflow design so the vendors are at least aware of this as a different use case. What do the rest of the white paper team think?)

(NOTE FROM NCJ: Hopefully between the contexts listed above, the consultation contexts ppt & discussion in the teleconference, and future discussion, we can all get to the same page on why these contexts have significant differences in workflow. The big question to me is do we want to make 1 big decision tree profile for all consultation, different trees for different contexts, or do the other authors have other suggestions? I must note that this is a big area where there is a gap in functionality for many vendor products, because these nuances have not been explained in white paper form previously. So while we might recoil from the complexity here, I think this is a key area of value of the white paper, but I am open to constructive criticism here.)

Use Case #2: Immunohistochemistry Positive Control Slides

Contributor assigned:

Creating digital copies of immunohistochemistry positive control slides to preclude the need for creating multiple positive control slides for distribution to pathologists

• Request for IHC stain processed as usual
• Only one IHC positive control run per batch
• IHC positive control slides imaged and saved to network folder
• Positive controls NOT distributed ($$$ savings)
• Glass IHC slides reviewed by pathologist but same positive control reviewed digitally by all pathologists for a given IHC (i.e. only a single cytokeratin positive control slide even if requested across 10 different patient samples)
  • Note that this imputes the need to have one slide associated with multiple cases. This will be beyond most models of slide/database ontologies, so vendors are advised to design systems that can reference control slides to multiple cases. (NOTE FROM NCJ: Question for group- how do IHC labs normally document where the positive control slides come from? Should databases know that positive control slide x came from patient-case y, but applies to slides for case-set z? How should this be modeled? How do you do this in your lab, Dr. Dash? We should discuss in the teleconference.)

To be refined and continued.

Use Case #3: Managing Digital Assets for Anatomic Pathology Clinical Workflows

Contributor assigned: JD Nolen

These workflows include for instance, Whole Slide Images for Primary Diagnosis. Creating digital copies of all glass slides for primary diagnosis

• Specimen collected and transported
• Specimen gross exam with possible digital imaging and annotation
• Specimen processing
• Glass slides produced as usual
• All glass slides fed into high volume automated digital scanner
• Scanner tags images requiring manual intervention
• Digital images deposited in network share, VNA, or PACS
• Interface message to LIS sent as each barcode read off slide
• Acknowledgment from LIS indicates case is valid and ready for association with digital slide assets
• Additional message sent when slide digitization completed
• Interface message sent every time slide viewed or annotated

To be refined and continued.

Use Case #4: Sharing and Cooperating on Gross Examination Images

Contributor assigned: Laurent Duval

TBD

Use Case #5: Incorporation of Legacy Digital Images for Use in APW

Contributor assigned: JD Nolen, Dan Rutz, Anil Parwani

TBD

Use Case #6: Image Analysis, Machine Learning and In Silico Workflows

Contributor assigned: Raj Dash

TBD

Use Case #7: Quality Control / Quality Assurance and Error Correction Workflows to Support Digital Pathology

Contributor assigned:Nicholas C. Jones

NCJ Note: Quality control and quality assurance are somewhat polysemous terms. See my ppt from shared dropbox area for some details. For clarity here, I will define 3 terms for use within the document. We can debate these if desired.

Quality Evaluation (QE): An assessment from a device, software algorithm, or person about the quality of a target physical or digital asset. (NCJ Note: My definition here, just to make the distinction between the common vernacular of QC - as in "he QC'd the image" versus the formal definition of quality control. I'm open to using a different phrase if desired.) Quality Control (QC): A system for verifying and maintaining a desired level of quality in an individual test or process. (From traditional CAP definition. My note: usually implies prospective processes, and to generate data.) Quality Assurance (QA): Systemic monitoring of quality control results and quality practice parameters to assure that all systems are functioning in a manner appropriate to excellence in health care delivery. (From traditional CAP definition.)

Use Case #8: Digital Pathology in Support of Clinical Conferences

Contributor assigned:Nicholas C. Jones

NCJ Question 1: How do we want to define clinical conferences? I would like to avoid the ambiguity of definitions like this. It may be easier to define through sub-types and examples, which is the approach I'll tentatively take here. But I would definitely like to discuss the tentative definition I'm suggesting here

Tentative definition of clinical conferences(for discussion): meetings of two or more clinicians and/or other clinical personnel, either in person, through tele-presence, or a combination of physical and tele-presence, to discuss and review clinical details of one or more patient's cases for prospective quality control purposes or retrospective quality assurance purposes, within the domain of clinical care for the patient. This does not include academic case studies in settings such as a professional association meeting. Clinical conferences are generally scheduled, as opposed to ad hoc processes, which would more commonly be considered consultations.

Categories include:

• Consensus conferences: common term used for routine conferences in which pathologists review one or more challenging or rare cases with colleagues prior to sign-out. Such conferences may or may not be noted in the final diagnosis of the report. Consensus conferences could occur at a multi-headed microscope, through tele-presence using static, dynamic, or whole-slide images, or any combination of imaging techniques.
  • In a multi-site/enterprise setting, consensus conference through digital pathology may be desirable as conferencing pathologists could be distributed geographically, and digital pathology supporting such conferences can allow for easy, efficient review of pathology images for group review.
  • In some circumstances, we could expect the desired combination of multi-headed microscope use along with digital pathology, such as in the scenario of prior consultation cases which were digitally archived needing to be reviewed along with glass slides at the multi-head. Alternatively, in a primary diagnosis setting, we may expect current cases would be reviewed through digital pathology systems being coupled with glass slides of prior cases (pulled prior to conference, but not scanned) being reviewed at a multi-head.
• Internal quality assurance conferences: Within a pathology department, certain cases may be retrospectively reviewed on an ad hoc and/or systematic basis. For instance, a cytopathology/histopathology QA conference might review cases in which the cytology and histology diagnoses disagreed to find root causes. (Such as issues of interpretation versus issues of sampling variation in cases involving fine needle aspiration biopsies and core biopsies.)
• Tumor boards or multidisciplinary conferences: In these settings, clinicians of multiple specialties review cases together. (Note we may make the distinction between tumor boards and multidisciplinary conferences as in some cases the patient may not have a tumor, yet the case can still warrant a multidisciplinary conference review.) In many instances, these conferences may occur outside the pathology department, so pathologists may traditionally show static images (or commonly a powerpoint with multiple static images), projections of a slide from a microscope, or whole slide images. Note that it is common for radiologists to use thin client viewers to present radiology images, so toggling to thin clients to present whole slide images is often considered highly desirable for such conferences. As the complexity of physical organization of clinicians and interconnected work grows, flexibility in presentation of such conferences is highly valuable to provide for clear, efficient presentation of case materials to ensure high quality conferences that minimize time-related switching costs in such conferences.

Technical notes:

• It is commonplace for slides to be imaged for the purposes of such conferences. Workflows may vary in a primary diagnosis scenario.
• Although such conferences may use "thick clients" for WSI viewers, it may be considered more commonplace to use thin clients for such scenarios. As such, in a thick-client primary diagnosis setting, the determination of a set of WSIs previously scanned may warrant upload of scans to a thin client hosting server, or transmission from a file store or from one thick client to another. (This latter scenario could occur if a hospital used a thin client system for internal client cases, and a separate thin client system for telepathology cases received outside the hospital network.)
• To provide inter-operable vendor support for such conferences, it may be desirable to support modeling of sets of requests for multiple patients to pathology for the purposes of such conferences. Common current practice may involve one or more parties communicating to pathology the request to review cases at the next conference by e-mail, but it may be preferable to allow for systematic requests (i.e. through the EHR) from parties like oncologists to pass from the EHR through interfaces to other systems (APLIS, PACS, and/or digital pathology systems) to assist in administration of such conferences. (Note from NCJ: Do we want to discuss what these orders might look like as HL7 messages? Does anyone know of examples of this?)
• Many digital pathology thin client systems allow for the organization of multiple cases, even for disparate patients, into sets or be indexed together in other forms, which can help assist in the process of organizing rapid, efficient, accurate switching between cases for multiple patients in these settings.
• While conferences are not generally tied to billing, from the perspective of cost accounting (saving multiple physicians time) they are extremely valuable. From the perspective of prospective clinical accuracy and providing constant feedback loops to and between clinicians, they are invaluable.

Use Case #9: Sub-contracting for special analyses on specimens

Contributor assigned: Gunter Haroske

Created as an increasing workload will be sub-contracted to special laboratories, e.g. for molecular analyses.

• Specimen collected and transported
• Specimen gross exam with possible digital imaging and annotation
• Specimen processing
• Glass slides produced as usual
• All glass slides fed into high volume automated digital scanner
• Scanner tags images requiring manual intervention
• Digital images deposited in network share, VNA, or PACS
• Interface message to LIS sent as each barcode read off slide
• Acknowledgment from LIS indicates case is valid and ready for association with digital slide assets
• Additional message sent when slide digitization completed
• Interface message sent every time slide viewed or annotated
• Case with all digital assets analysed by the pathologist
• Selection of relevant assets/slides/blocks for consultation or sub-contracting
• Mailing of relevant assets/slides/blocks to external lab
• AP reporting (preliminary)
• Mailing of the (preliminary) report
• Receiving results of consultation / sub-contracted tests
• critical reflexion
• AP reporting (synoptic, final)
• Mailing of the final synoptic report.

Use Case #10: Image Registration Functions

Contributor: Nicholas C. Jones

Image registration involves the creation of a cross-image coordinate system of multiple images for the same object(s) or "scene."

• Example for explanation: since many liken WSI to GoogleMaps, let's use GoogleMaps (used here as GoogleMaps + Google Earth functionality) as the example for analogy. Both WSIs and GoogleMaps involve the creation of numerous individual images through an optical and camera system. WSIs involve capturing numerous micographs (each a mosaic of pixels) in a mosaic pattern of micrographs, stitched together, while GoogleMaps took numerous satellite images of the Earth and stitched them together. But GoogleMaps also includes complex image registration algorithms, where images from space were correlated to images of the earth (various types of maps) to create a more complex but more accurate model of the Earth. Together, they allow for easier use of more complex path creation, but also to toggle in a location between a "street map view" and the satellite view of the same location. Image registration functions are used in other areas of medical imaging to correlate multiple modalities of imaging for the same organ(s).

• 10.a. Image registration for swapping to correlated position on different slide from a single block. This feature allows for the user to swap to another slide from the block (i.e. going from A1-1 H&E to A1-2 CK8) to the analogous location on the new slide without the need to visually search for it. Used in a larger series, it allows for oscillating between multiple slides in the same location (ie. A1-1 H&E, A1-2 CK8, A1-3 CK 14, A1-4 ER, A1-5 PR, A1-6 her2... etc.)
• 10.b. Image registration for correlated multi-slide view from a single block. This allows for simultaneous presentation of multiple slides at once off the same block by allowing the user to control the location of one slide and having the system move the view of the other slides to the corresponding locations. (Note: "manual" control/locking of multi-slide views is basically similar here but is not using an automated image registration system to create the alignment of views.)

10.c. Image registration beyond just a series of serial sections of a block have been researched, and although these are (to our knowledge) not currently in clinical use, further possibilities have been documented in research environments. These include creating 3D virtualizations of blocks by registering large series of WSIs, or registration of multiple blocks with gross images and through those to other images such as those from radiology. (NCJ Note: if we want to include this, do you want references to work from Dr. Yagi and others?) Vendors, academics, and standards may want to consider these more complex image registration as future possibilities and thus design ontologies and systems that are extensible enough to warrant these use cases in the future. (NOTE FROM NCJ: Does anyone know of any groups using image registration from WSIs/blocks to gross images in a clinical environment? If so, we should separate that out from the "future use" section.)

Note that in the case of doing image registration in multiple slides, this may be considered a loose definition of image registration, as technically the different slides could be considered different "scenes" as a different object is being imaged. But since the slides are different serial sections cut off an FFPE block, it is ideally very close to the same scene (with ~5 um or even less between serial sections). Still, that difference between the slides makes these image registration processes more challenging, as variations between the levels off the block will show small variation in shape, size, and features from other slides.

• It is important for the success of such algorithms for systems to pass identifying metadata not just on unique slide identifiers, but also slide section thickness (such as 5 um), block identifiers, and tissue cut order (i.e. A slide called "A1-7 H&E" on the label may mean the 7th slide off the block "A1," but its unique identifier may not itself indicate that this slide is the 7th off the block, especially if the APLIS unique identifiers are created automatically in the database by order of their creation as a database asset rather than physical cut order.)
• Furthermore, tools for manual intervention to change orientation of tissue may be required to aid these algorithms. For instance, original H&Es may have been cut with epidermis facing in one direction, while the subsequent cuts for IHC may have a different orientation. Tools to permanently rotate images in multiples of 90 degrees, or temporarily rotate images to aid in image distribution may assist in these processes and be part of active troubleshooting, as image registration may be done on-demand by technicians prior to passing WSIs onto pathologists.

Use Case #11: Digital Pathology in Support of Intraoperative Procedures

Contributor assigned:

Such as frozen section interpretation and rapid fine needle specimen adequacy assessment.