NCI Workshop on Digital Pathology Imaging (DPI) in Cancer Clinical Trials and Research

Hosted by: DCTD’s Cancer Diagnosis Program

March 6-7, 2024

Workshop Recordings: Day 1 , Day 2

On this page:

• Workshop Overview
• Workshop Highlights and Essential Conclusions
• Key Ideas from Presentations and Discussions
• Speakers

Workshop Overview

Audience: Of the 392 attendees, 32% were from the U.S. government, and 68% were non-government

• NCI — 25% NCI and non-NCI government — 7%
• academic — 46%
• industry — 17%
• other — 5%

The diversity of professional backgrounds showcased the interdisciplinary nature of digital pathology (DP).

pathology/DP/biology — 38%
oncology/clinical oncology — 10%
data science/IT/(bio)informatics — 27%
radiology/medical imaging/imaging analysis — 8%
artificial intelligence/machine learning — 6%
biobanking — 5%

Focus: Expanding the roles of DPI in translational cancer research, biomarker studies, clinical trials, and pharmaceutical development.

Objectives: Understanding the specific needs for DPI by investigators and biospecimen banks and successfully integrating DPI into cancer clinical trials.

Presentation Topics: Successes and challenges in digital and computational pathology, including hardware and software, image acquisition, validation, storage, data management, intellectual property, and public-private partnerships.

Discussion: Addressing challenges posed by the current lack of standardized approaches for DPI utilization in clinical trials and biobanking and propose potential solutions.

Event Organizers:
Drs. Hala Makhlouf and Irina Lubensky

Advisory Committee
Mark Watson (NCTN Group Banking Committee [GBC], Washington University, St. Louis)
William Richards (GBC, Harvard University)
Shakeel Virk (GBC, Queen’s University, Canada)
Lyndsay Harris (Associate Director, CDP, DCTD)
Keyvan Farahani (National Heart. Lung, and Blood Insitute, NIH)
Mathew Hanna, (Memorial Sloan Kettering Cancer Center)

For questions about the meeting or to request copies of slide presentations, contact Dr. Hala Makhlouf (hala.makhlouf@nih.gov) or the Cancer Diagnosis Program (NCICDPNews@mail.nih.gov).

Workshop Highlights and Essential Conclusions (Recordings: Day 1 , Day 2 )

DP Infrastructure

A comprehensive infrastructure beyond scanners for effective DP implementation is needed.

Standardization and Software

DICOM is the DP standard, and support from scanner vendors and software solutions for image conversion are important.

Imaging and Analysis

Microns per pixel (mpp) is more important than magnification level for scanning slides, and Standard Operating Procedures (SOPs), imaging performance criteria, and access governance are crucial for DPI in clinical trials.

Data De-identification and AI

Open standards in data de-identification and DPI archives for AI analysis and research enhancement are important.

Platforms and Access

Attendees outlined the NCI Imaging Data Commons (IDC) as a platform for cancer pathomics research and the differences in access policies between IDC and the Cancer Imaging Archive (TCIA).

Challenges and Innovations

Examples of Advancements in AI and DPI Technologies:

• Successes in AI integration in clinical research (e.g., studies enabling real-time diagnostic capabilities during surgeries) and enhances in precision in oncology treatments
• FDA approval of the first AI-assisted digital cytology system for Pap test screening
• Enhancements in the value of clinical trials by improving outcomes and treatment strategies (e.g., NCI-MATCH)

Concerns and Challenges Highlight the Following Needs:

• Interoperability and Standardization: Integrating diverse systems for seamless data sharing and analysis across different platforms and scanners
• Resource Intensity and Infrastructure: High-quality, anonymized image files in standardized formats and funding for long-term storage and distribution
• Data Privacy and Security: Effective de-identification methods that do not limit data utility (encompasses ethical considerations and regulatory compliance, especially in handling personal health information (PHI))

Key Ideas from Presentations and Discussions

Presentation Title and Key Ideas	Speaker
Day 1: March 6, 2024
Session 1: Digital Pathology Imaging (DPI) Landscape: Insights and Lessons Learned	Moderator: Ping Guan
Precision Oncologic Pathology: The Impact of Disruptive Technologies Next-generation tissue-based biomarkers are likely to be identified by use of large, well-curated datasets. Next-generation pathologic analyses require platforms that can characterize the co-expression of key molecules on specific cellular subsets, in situ and in spatial context.	Dr. George Netto
Digital Cytology: Applications and Lessons Learned Given advances in whole-slide imaging (WSI) technology coupled with AI, digital cytology no longer trails other fields. The FDA recently cleared the first digital cytology system for AI-assisted Pap test screening.	Dr. Liron Pantanowitz
Digital Pathology to Enhance the Value of Clinical Trials: The NCI-MATCH Experience The NCI-MATCH precision medicine clinical trial demonstrated the benefits and relevance of DPI. SOPs, imaging performance criteria, governance of access, and investment in DP are needed. Investigators should be aware of the resource intensity of infrastructure and logistical considerations in trials.	Dr. Stanley Hamilton
Digital Image Management: Lessons Learned in Radiology Consider image storage, normalization, and annotation. Open standards can promote interoperability. Syntactic interoperability equates with the technical ability to transmit, store, and retrieve images. Semantic interoperability enables data discovery and re-use.	Dr. Kenneth Wang
The Cancer Imaging Archive: Perspectives and the Path Forward TCIA provides de-identification and curation services. Data types range from radiology, digital histopathology, associated clinical datasets, and analytical results. TCIA links to platforms and code to facilitate AI analysis.	Dr. Lalitha K. Shankar
Q&A and Panel Discussion Annotated slides and images from NCI-MATCH will need further processing before they can be widely available. Hosting NCI-MATCH data on TCIA is being considered. TCIA is valuable for data sharing that NCTN leadership supports. It includes some studies and datasets with multiple timepoints, but not many with digital histopathology data, and it can host any digital IHC data for NCTN studies. Telecytology and computer-aided detection has improved turnaround time on pap smear results, but disparities remain between processing times in certain geographical locations, likely due to workforce shortages and logistics. NCI does not currently endorse a commercially available de-identification method or anonymizing tool; TCIA accepts de-identified data from various systems to prevent limiting the datasets that can be hosted and shared. TCIA is open to using sequestered spaces for closed access to certain NCTN datasets (e.g., ETCTN); however, Imaging and Radiation Oncology Cores (U24), which support the NCTN groups and have strong connections with their statistical centers, typically conduct the analyses.
Session 2: Exploring Investigators’ Digital Pathology Imaging Needs and the Science of Artificial Intelligence (AI)-Driven Research	Moderator: Miguel Ossandon
High-Quality Data to Drive Future Thinking in AI Consistent data acquisition and imaging formats are needed to achieve broad analysis spanning multiple trials. Consistency allows for training of foundational models (e.g., cell segmentation algorithms) and using inter-trial data.	Dr. Amber Simpson
Developing Multi-Modal Tissue Imaging Approaches to Guide Personalized Therapy Quantitative molecular pathology imaging requires new tools and approaches to handle vast amounts of information. Multi-modal tissue imaging for precision medicine in oncology needs innovative technologies that integrate molecular precision with human intelligibility.	Dr. Sandro Santagata
Whole Slide Imaging Biomarkers and Foundation Models WSI data provide valuable insights into molecular details and tumor immune system interactions. Although traditional supervised learning methods remain valuable, exploring weak learning and self-supervised techniques may yield more robust and easily trained biomarkers.	Dr. Joel Saltz
Real-Time Cancer Pathology Assessment via Artificial Intelligence AI-empowered pathology imaging analyses facilitated real-time brain cancer diagnosis during surgery. Multi-disciplinary research is required to address technical and implementation challenges.	Dr. Kun-Hsing Yu
AI and Digital Pathology: Validation on Completed Clinical Trials Computational analytics with routine imaging could help address questions in precision medicine (e.g., prognosis and predicting response to therapy). Intentional and focused research on interpretable computational based biomarkers is needed.	Dr. Anant Madabhushi
Q&A and Panel Discussion Multiplex immunofluorescence data are training AI models to predict stain distribution in H&Es and are using the molecular labels to inform H&E algorithms. Molecular labels from fusing antibody data are powerful and informative. A tremendous amount of data comes from H&E, even without phenotyping. A prognostic role has been demonstrated in several cancer types (e.g., with TILs). Fusing different data types to use with AI models has potential to push the simulation limits with H&E. More richly annotated data will allow the scientific community to move this forward. Widely available data are important, but there is value in having high-quality/high-performance compute closed systems available only to trusted researchers. Data governance allows linking to administrative data including long-term patient outcomes from epidemiological studies. Multimodal learning can incorporate additional molecular information into prediction models. With more genetic information, improved AI models that combine gene expression and pathology data will be helpful. Prognostic predictive LDTs using DP exist, but return on investment may be a barrier to clinical use. Institutional support and reimbursement opportunities may address this, and the changing regulatory landscape can encourage transition from research to clinical workflows.
Session 3: NCI Clinical Trials Biobanks Experience with Digital Pathology Imaging: Successes and Challenges	Moderator: Irina Lubensky
Challenges and Solutions for Leveraging Digital Pathology in National Clinical Trials Network (NCTN) Develop systems for easy delivery of high-quality, anonymized image files in a standard format from diverse clinical sites, enabling focused and broad scientific use. Secure funding or partnerships for prioritized slide scanning, long-term storage, and distribution. Create a strategy for the careful collection and storage of associated patient, specimen, scanning, and annotation data for correlative studies.	Dr. Mark Watson
The Early Phase and Experimental Therapeutics Trials (EET) Biobank Digital Pathology Experience The Biobank collaborates closely with EET investigators, providing support to enhance the use of DPIs. Research studies connect with expert pathologists to aid in DPI analyses.	Dr. Nilsa Ramirez
Digital Pathology Insights, Canadian Cancer Trials Group Perspective Implementing DP in biobanks requires more than just scanners. Industry should embrace a standardized single-slide format such as DICOM to mitigate interoperability issues, especially in image analysis and AI.	Dr. Shakeel Virk
Collaboration with the NCTN Clinical Trials Statistical Centers to Assess the Impact of Digital Pathology Research Inventories of available NCTN trial DP data need to be constructed to maximize data use. Data use agreements are required to merge existing DP data and clinical trial treatment and outcome data for new analyses. Simplify access to the image and a small defined set of clinical trial data.	Dr. William (Bill) Barlow
Q&A and Panel Discussion Differences across jurisdictions in human data sharing legal frameworks require compliance considerations. Express informed consent must be obtained from donors for the collection, storage, and sharing of samples, and must comply with the specific requirements for each geographical jurisdiction. Ethics committees like IRBs ensure that biobanks are regulatorily compliant and adhere to all applicable legal standards and guidelines. Data security and privacy agreements, as well as data and materials sharing agreements exist. PHI must be masked, and the image must retain a unique identifier to maintain provenance. Many vendors provide de-identification tools for viewing, sharing, and annotating, but this can be problematic for image analysis tools that require specific formats. Reviews of biospecimen requests use a standardized process that evaluates scientific merit, ethical considerations, and regulatory compliance. Some banks may use a tiered fee structure. Intellectual property matters are addressed in the materials transfer agreement. Making NCTN WS images publicly available requires leveraging resources to protect the integrity of the trial, PHI, and the network. Images that are not centrally sourced should include metadata to describe how the image was captured (e.g., z-plane, magnification, etc.).
Session 4: Digital Pathology Imaging Standardization, De-identification, and Validation	Moderator: Keyvan Farahani
DICOM in Pathology as a Standard Interoperability is essential for scalability, and DICOM format can/should be universally used. Most vendors now support the format in new scanners. The most popular libraries like OpenSlide and BioFormats already support the format.	Dr. David Clunie
Open-Source Tools for WSI Image Deidentification Slides are often scanned using proprietary image formats making a universal image Deidentification (DeID) pipeline difficult. Open-source WSI readers exist for popular formats, facilitating image DeID. Tools that avoid image-recompression are important to avoid introducing artifacts.	Dr. David Gutman
Digital Pathology and AI Validation Validating DP for primary diagnosis involves adhering to guidelines and good practice statements, such as those provided by the College of American Pathologists. This ensures that lab-developed tests are rigorously tested and validated. The research phase and model development are critical steps, ensuring that the DP tools and AI algorithms meet the necessary standards for accuracy and reliability in clinical settings.	Dr. Matthew Hanna
Q&A and Panel Discussion Documenting “Safe”/“Unsafe” metadata attributes for AnyPI, SVS, and TIFF images is available in open source regular expression and validation rules, but it could be simplified and linked on an NCI wiki for ease of reference. DeID is context-specific for different use cases; an appropriate blanket certification procedure would be difficult to develop, especially due to the complexity and commercial costs. Testing and evaluation frameworks are useful when considering solutions. Guidelines or best practices like those developed through NCI’s MIDI Challenge may be used. When converting proprietary vendor formats to DICOM, transforming without losing pixel quality is essential. Commercial, shovel-ready tools and open-source tools of good quality are available, but capabilities vary, and tools should be evaluated based on use case. DICOM defines a standard as an interface boundary, which is a protocol to a system. Within the system, a user could separate the metadata from the bulk data because the macro and label data are stored in separate instances. Metadata and image pixel data are traditionally grouped, but DICOM separates access to metadata. The MIDI project is limited to radiology due to the available radiological dataset with synthetic PHI; however, a large collection of DP images with synthetic PHI is available and expanding the activities to DP may occur in the future.
NCI Funding Initiatives for Digital Pathology Imaging Projects
Funding Announcements: Challenges and Opportunities for Digital Pathology	Miguel Ossandon
Day 2: March 7, 2024
Session 5: Digital Pathology Imaging Hardware and Software: Challenges and Solutions	Moderator: Brian Sorg
Overview of the Hardware Landscape The performance of WSI scanners relies on the quality of their components and features like tissue finding accuracy and additional focus points, slide characteristics, and image loading capabilities. Automated scanners offer higher throughput and efficiency. Scanners typically come with magnification of 20x and 40x objectives with resolution of 0.5 microns and 0.25 microns per pixel, respectively. The 20x objective is effective when used for routine viewing of H&E and IHC slides, and the 40x objective is used for high accuracy on cytology, hematology, or renal specimens. Challenges include substantial infrastructure investment and slide storage and management,	Dr. George Yousef
Setting Up Scanning Operations: Costs and Quality Considerations Infrastructure building should be based on identified resources and expected research volume. Additional middleware and software solutions will need to be scaled to research needs. Integration of AI tools for research should follow DP infrastructure building.	Dr. Orly Ardon
From Whole-Slide Imaging to No-Slide Imaging H&E-stained slides are not the only ground truth. Advantages of no-slide imaging for biorepository research: Faster, non-destructive, novel structure, and contrast	Dr. Richard M. Levenson
In-Line Quality Control for Whole-Slide Imaging Implementing rigorous in-line quality control protocols is crucial for ensuring the integrity of WSI. QC should be integrated into the workflow efficiently to minimize disruptions. Not all artifacts need scan rejection.	Dr. Mark D. Zarella
“All In” on a DICOM-Centric Pathology Infrastructure Without standards like DICOM, siloed projects and wasted time, effort, space, etc. regarding sharing, retrieval, and analysis may occur. DICOMweb is hard to learn and has limited support for DPI in some tools.	Dr. Steven Hart
Software Tools to Accelerate AI Development for Digital Pathology Annotation software should support multiple annotators and remote access. Development of AI can be accelerated using dptools - a software package that insulates developers from the technical challenges of handling WSIs. These tools support the development of AI models and their deployment in the clinic.	Dr. Anne Martel
Q&A and Panel Discussion A cost analysis for making capital investments in DP revealed that a significant estimate of ROI from the DPA calculator may be useful, but there are always unknowns. Detailing the benefits of DP may be more persuasive in proposed business plans. The cost of adoption may be offset by maintaining a competitive edge in the field, and patient benefits should be emphasized. Some countries address instrument obsolescence by leasing, but IT depreciation is an issue for many fields. Vendor involvement with trade-ins or repurposing scanners that reach the end of their useful life is necessary. When choosing a scanner, use case and cost-effectiveness must be considered. An integrated, widely compatible system is important. Hardware, scanners, and storage can be sourced from different vendors; however, compatibility is not always simple to achieve. Practitioners and interventionalists are the major customers of digital images; therefore, acceptance and adoption of slide-free technology is made easier by the natural separation of pathologists from image generation. Deployment costs, regulatory considerations, and training are the major considerations for acceptance of the technology. DPI is changing the pathology landscape similar to how NGS changed genomics, but it is less disruptive when considering equipment cost and the existing workflows for sample processing.
Session 6: Digital Pathology Imaging Data Management and Storage	Moderator: Rodrigo Chuaqui
NCI Imaging Data Commons as a Platform for Cancer Pathomics Research NCI IDC is a powerful platform for cancer research, offering a rich repository of harmonized, well-documented, and easily accessible imaging data, plus advanced tools for data analysis. By adopting the DICOM standard, the IDC ensures that images, annotations, and related data are consistent and compatible across systems, which is essential for high-quality research. Cloud technology offers scalable resources for collaborative and reproducible research.	Dr. Andrey Fedorov
Research Pipelines for Digital Pathology Data at Memorial Sloan Kettering Cancer Center MSKCC has developed sophisticated research pipelines tailored for DP data, internal metadata whitelist for SVS, and De-ID.	Luke Geneslaw
Enterprise Platforms for Image Management and Analysis: Requirements and Solutions Integrating cutting-edge data science tools, such as advanced nuclei segmentation with OMERO, enhances research efficiency and data analysis. Enterprise data science platforms are vital for cancer research.	Dr. George Zaki
Image Data Warehouse Solution for Easy Navigation, Annotation, Analysis and Sharing of Biospecimen Related Research Data New tools such as DP and computational pathology aided by AI are available for clinical and research applications. Pathologists, vendors, and researchers must continue to collaborate to build innovative tools that will enhance the biospecimen workflow of the future.	Dr. Anil Parwani
Q&A and Panel Discussion The cost structure for researchers accessing data via the MSKCC infrastructure involves the staffing to build the tools that provides this data, which is best handled as an enterprise cost. A large amount of image storage requires a high compute. Costs vary based on study size; for smaller studies, grant funding for storage may be appropriate, but large scale requires enterprise funding. A simple but important way to reduce storage costs is to avoid duplicating slide images. Infrequently accessed data can be stored in lower-performance, lower-cost archive solutions. DPI and AI analysis needs are growing, even for basic and translational researchers. These data are useful for building predictive models with imaging data in basic sciences. Open-source tools are available that provide image management, but they may not interface with a clinical system. The primary scanning vendors that provide interfaces with LIM systems are appropriate for research settings. Using data commons for data sharing should be considered as early as possible in a study. The IDC framework can help as soon as samples are available, providing tools for data harmonization, then commercially available tools to explore DICOM infrastructure can make images readily available. Images can be integrated with open-source users for annotation. Using PathPresenter, users can download data to perform local analysis. Similarly to specimen distribution under CHTN, users will sign a DUA to access images for research.
Session 7: Integrating Digital Pathology Imaging into the Clinical Trial Enterprise: Challenges and Solutions	Moderator: Tracy Lively
NCI-MATCH: The First Systematic Attempt to Explore Homogeneity and Heterogeneity of Response in Different Cancer Types When Targeting Common Targets Pathomics provides an impartial approach to identifying potential predictors of treatment response and resistance that could lead to better-tailored therapies.	Dr. Keith T. Flaherty
Digital Pathology in Pediatric Sarcoma — an Opportunity to Implement Advanced Analytics to Patient Diagnosis and Stratification Applying DP in pediatric sarcoma involves using advanced AI tools (Necrosis-Digital Analysis Software (N-DAS) and Digital Correlation of MRI (D-COR)) to assess chemotherapy responses. This integration 1.) aids in developing biomarkers and enhancing osteosarcoma analysis accuracy; 2.) addresses challenges in multi-site studies, diverse imaging devices, and data storage essential for advancing AI and DPI in sarcoma research.	Dr. Patrick Leavey
Strategies and Tools for Developing AI Models into Deployable Clinical Biomarkers Consider the clinical application and thresholding plan (outlines the criteria for interpreting AI outputs) and the intended deployment framework and workflow integration. Prioritize bias identification and mitigation from an early stage.	Dr. James Dolezal
Q&A and Panel Discussion Pathology imaging analysis can soon be leveraged to prioritize samples that are likely to have actionable mutations, and sensitivity optimized confirmatory NGS testing is possible. In lower-resource settings, a highly specific pathology test could decide treatment based on high likelihood of actionable mutations. The level of phenotypic association between molecular alterations and histologic manifestations will differ based on use case. When a trial is designed to validate biomarker use, the trial design could include therapy escalation for high-risk biomarker validation. A case study demonstrated significant cost savings when using a biomarker as a first stage test to predict HPV in head and neck cancer without confirmatory IHC or PCR testing. One potential use case for biomarkers is as a first line screening for clinical trials in populations with rare molecular alterations that are potentially druggable.
Session 8: Exploring the Issues of Digital Pathology Imaging: Intellectual Property and Academia-Industry Partnerships	Moderator: Lokesh Agarwal
A Brief Overview of IP Issues Involved in Uses and Sharing of Digital Pathology Images Proper management of IP rights is crucial in DPI. It is important to maintain good records of all agreements pertaining to DPIs, understand usage permissions and restrictions, and enforce your rights when necessary to ensure IP compliance. Always acknowledge the source of DPIs in publications, and thoroughly review software licenses when using DPIs for image analysis.	Dr. Lynne Huang
Opportunities with Public-Private Partnerships Public-private partnerships in DP are critical for tackling major health challenges and are instrumental in driving biomedical innovation.	Dr. Stacey Adam
Q&A and Panel Discussion Medical images are copyrightable, but the copyright owner may not always be obvious. When commercially available scanners are used, institutions are most often the owner. When prototype scanners are used, a collaboration may generate the images between the parties, and the scanner developer may have certain rights. Data owners must be aware/made aware of the purpose for which the data is being used (i.e., to train AI models). In a collaboration, the owner may set conditions for the model that is created by using their data. Licensing agreements will determine whether data owners share ownership or receive royalties from the model, especially if it will be marketed. The relationship between a pathologist and the institution can determine whether the pathologist owns the models that are trained using their diagnostic reports or image annotations. Often, IP generated by an employee is owned by the institution; in contracted cases it would be determined by the contract or license between the parties. FNIH convenes an expert panel on their cancer steering committee once a year to debate areas to pursue, and for DP, they identified the benefit of building a collaborative to do standardization work rather than pursuing specific biomarkers. Ensuring harmonized data collection and alignment of standards and data evaluation is not being addressed by one entity, and FNIH encourages collaboration to promote accessibility.

Speakers

Stacey Adam, PhD, is the Associate Vice President at the FNIH, leading many public-private partnerships, such as Accelerating COVID-19 Therapeutic Interventions and Vaccines (ACTIV); the Biomarkers Consortium (Cancer and Metabolic Disorders Steering Committees) and their projects; Accelerating Medicines Partnerships (AMPs)-Common Metabolic Diseases, Heart Failure, and Parkinson’s Disease, Partnership for Accelerating Cancer Therapies (PACT); and the Lung Master protocol (Lung-MAP) clinical trial. Prior to FNIH, Dr. Adam was a Manager at Deloitte Consulting in the Federal Life Sciences and Healthcare Strategy practice where she supported many federal and non-profit client projects.

Orly Ardon, PhD, MBA, is the Director of Digital Pathology Operations and an assistant member at the Department of Pathology and Laboratory Medicine at Memorial Sloan Kettering Cancer Center. Before joining MSKCC, she led the development of novel computer assisted diagnostic tools and the expansion of digital pathology collaborative initiatives at ARUP Laboratories in Salt Lake City, Utah. Dr. Ardon’s work in digital pathology is centered on operational aspects of the digital technologies and the clinical implementation of laboratory automation. Her interests also include process improvements and healthcare economics. She a Board Member at the Digital Pathology Association (DPA).

William Barlow, PhD, is a Senior Biostatistician at Cancer Research and Biostatistics and a Research Professor in the Department of Biostatistics at the University of Washington, Seattle, WA. He is the primary statistician for all breast cancer clinical trials conducted by SWOG NCTN Group and a statistical representative to the NCTN Group Banking Committee. His primary interests are evaluation of breast cancer screening, breast cancer treatment efficacy, and the predictive effect of markers on cancer treatment. Dr. Barlow has also conducted research on vaccines, backpain, ophthalmology and case-cohort designs.

David Clunie, MBBS, is a radiologist, medical informaticist, DICOM open-source software author, and editor of the DICOM standard. He was formerly the co-chair of the IHE Radiology Technical Committee and industry co-chairman of the DICOM Standards Committee, a chairman of several of the DICOM working groups, including structured reporting, digital x-ray, compression, interchange media, base standard, display, mammography, clinical trials, preclinical small animal imaging, digital pathology, and conformance. He serves as an NCI SME contractor as well as being a sub-contractor to BWH in the NCI Imaging Data Commons (IDC) program.

James Dolezal, MD, is a medical oncologist and computational scientist at the University of Chicago. He specializes in the integration of clinical oncology and digital pathology through AI. His research is centered on the development of deep learning biomarkers for upper aerodigestive malignancies, along with methodological advancements to improve safety and reliability of AI models designed for clinical settings. He spearheads an open-source DPI software initiative that focuses on augmenting the accessibility and reliability of building DPI AI models and is involved in the development of algorithms and software tools that support the conversion of these models into actionable clinical biomarkers.

Andrey Fedorov, PhD, is a researcher at Brigham and Women's Hospital (BWH) and Associate Professor of Radiology at Harvard Medical School. He is one of the leads of the team tasked with building National Cancer Institute Imaging Data Commons (IDC). A Ph.D. computer scientist by training, Andrey spent past ~15 years at the BWH Surgical Planning Lab working on translation and evaluation of image computing tools in clinical research applications. He is dedicated to developing infrastructure and best practices to help imaging researchers improve transparency of their studies, simplify data sharing, and make their analyses more easily accessible and reproducible by others.

Keith Flaherty, MD, is the Director of Clinical Research at the MGH Cancer Center, and Professor of Medicine at Harvard Medical School. He is the principal investigator of the NCI MATCH trial, the first NCI-sponsored trial assigning patients to targeted therapy independent of tumor type based on DNA sequencing detection of oncogenes. He contributed to the treatment of melanoma by establishing the efficacy of BRAF, MEK and combined BRAF/MEK inhibition in patients with metastatic melanoma. Dr. Flaherty joined the NCI Board of Scientific Advisors in 2018 and AACR Board of Directors in 2019. He serves as editor-in-chief of Clinical Cancer Research.

Luke Geneslaw, MBA, is a Senior Product Manager in the Department of Pathology and Laboratory Medicine at Memorial Sloan Kettering Cancer Center. He has developed de-identification pipelines which repurpose digital pathology data for research and educational uses at scale. His recent focus is in building applications and integrating decision support tools fostering adoption of digital and computational pathology in clinical settings.

David Gutman, MD, PhD, is an Associate Professor in the Department of Pathology at Emory University School of Medicine. His research is focused on developing innovative tools for analyzing large imaging datasets focusing on whole slide images in pathology. He has been instrumental in creating the Digital Slide Archive platform and HistomicsTK, an open-source suite for managing and analyzing large-scale image sets. These tools have been used to investigate aspects of cancer biology research, particularly from TCGA related data sets. He also contributed to developing open-source tools for image de-identification, further enhancing the utility and security of whole slide imaging datasets.

Stanley R. Hamilton, MD, FCAP, is a digestive system and molecular pathologist. He is a Professor and Chair of the Department of Pathology at the City of Hope National Medical Center and Comprehensive Cancer Center. He is also the Director of the Clinical Trials Specimen Qualification Laboratory and the Research Pathology Services Shared Resource. His current research interests focus on development and clinical applications of novel biomarkers for precision oncology clinical trials. The clinical laboratories at City of Hope support the conduct of clinical trials through biospecimen qualification, regulatory-compliant laboratory testing, translational research studies, and use of digital pathology.

Matthew G. Hanna, MD, is the Director of Digital Pathology Informatics at Memorial Sloan Kettering Cancer Center. He is a pathologist with expertise in breast pathology, informatics, digital & computational pathology. His clinical interests include breast pathology, informatics, digital pathology, image analysis, machine learning, clinical operations/implementation, and decision support tools. Dr Hanna serves as a board member of the Digital Pathology Association and NY Pathological Society. He also actively contributes to the CAP as the Vice Chair of the Artificial Intelligence Committee and member of the Informatics Committees.

Lyndsay Harris, MD, is a medical oncologist who joined the NCI in 2016 after a 30-year career in breast cancer research and translational science. As Associate Director of the Cancer Diagnosis Program (CDP) her role is to support the development of robust prognostic and therapeutic biomarkers. The CDP works closely with the Cancer Therapy Evaluation Program (CTEP) to implement biomarkers into several Precision Medicine trials including the NCI-MATCH (Molecular Analysis for Therapy Choice) trial and ComboMATCH. Dr. Harris manages the MDNet assay network to provide molecular assays services on three new Precision Medicine trials: iMATCH, ComboMATCH, and MyeloMATCH.

Steven Hart, PhD, is an Associate Professor and Senior Associate Consultant-AI at Mayo Clinic. He has a PhD in Pharmacology from the University of Kansas Medical Center and has spent his career developing and implementing bioinformatics software. He leads much of the data strategy, computational infrastructure, and operational oversight for the Division of Computational Pathology and AI. He is also a co-Chair of the Association for Pathology Informatics’ Technical Standard Committee.

Lynne Huang, PhD, JD, is a Senior Intellectual Property Adviser at the NCI Division of Cancer Treatment and Diagnosis (DCTD). She takes the lead in reviewing and negotiating a variety of transactional or collaborative agreements for meeting the overall needs of CTEP clinical programs, including Precision Medicine Initiatives, and provides expert guidance and advice as necessary for IP and data protection. Dr. Huang works closely with multiple branches within DCTD, extramural clinical networks/sites, NCI contractors, and NCI pharmaceutical collaborators, to ensure that the terms of any agreements/ contracts reflect current processes of programs and policy requirements.

Patrick Leavey, MD, is a Professor of Pediatrics, and the Associate Vice Chair for Research Operations and Interim Chief of Pediatric Hematology/Oncology at UT Southwestern Medical Center, Dallas Texas. Dr. Leavey led the most recent Children’s Oncology Group (COG) Phase III study for patients with non-metastatic Ewing Sarcoma and is a funded translational investigator navigating the use of advanced analytics and digital imaging to improve treatments for children with bone and soft tissue sarcoma.

Richard Levenson, MD, FCAP, is Professor and Vice Chair for Strategic Technologies, Department of Pathology and Laboratory Medicine, UC Davis Health. He received his MD at University of Michigan and pathology training at Washington University, followed by a cancer research fellowship at Univ. of Rochester and faculty positions at Duke and Carnegie Mellon. He then joined Cambridge Research & Instrumentation, Inc., becoming VP of Research before assuming his present position at UC Davis. He helped develop multispectral microscopy and small-animal imaging systems, birefringence microscopy, multiplexed ion-beam imaging (MIBI), and slide-free as well as enhanced-content microscopy approaches, and is an inventor on some 10 patents. He is section editor for Archives of Pathology and is on the editorial board of Lab. Invest. and AJP. Regrettably, he also taught pigeons histopathology and radiology. He is a recipient of the 2018 UC Davis Chancellor’s Innovator of the Year award and is a Fellow of SPIE.

Irina A. Lubensky, MD, is the Chief of Pathology Investigations and Resources Branch (PIRB) at the Cancer Diagnosis Program (CDP), DCTD, NCI. Prior to joining CDP, she served as a Surgical Pathologist and Chief of the Hereditary Cancer Syndrome Unit at the NCI Laboratory of Pathology and as a translational researcher at the Surgical Neurology Branch, NINDS, NIH. As PIRB Chief, she oversees NCI multi-institutional cooperative agreement grants for biospecimen banking to support cancer research including Cooperative Human Tissue Network (CHTN) and NCI Clinical Trials Network Biospecimen Banks (NCTN Banks), as well as the NCI Specimen Resource Locator.

Anant Madabhushi, PhD, is the Robert W Woodruff Professor of Biomedical Engineering, and is on faculty at the Departments of Pathology, Biomedical Informatics, Urology, Radiation Oncology, Radiology and Imaging Sciences, Global Health and Computer and Information Sciences at Emory University. He is also a Research Career Scientist at the Atlanta Veterans Administration Medical Center. He has more than 200 patents in the areas of artificial intelligence, radiomics, medical image analysis, computer-aided diagnosis, and computer vision. He received the 2017 IEEE Engineering in Medicine and Biology Society (EMBS) award for achievements in computational imaging and digital pathology.

Hala R. Makhlouf, MD, PhD, is a pathologist and Program Director at the Cancer Diagnosis Program (CDP), DCTD, NCI. She manages multi-institutional cooperative agreement grants for NCI Clinical Trials Network Biospecimen Banks (NCTN Banks) and Early-Phase and Experimental Therapeutic Clinical Trials Biospecimen Bank (EET Bank). She collaborated with NCTN Biobanks to create guidelines for DPI within NCTN trials and provides pathology expertise through evaluating and annotating DPI of biopsy and resection specimens for DCTD investigators. Prior to joining CDP, Dr. Makhlouf served as Chief of the Division of Hepatic Pathology and as Acting Chairman of the Gastrointestinal and Hepatic Pathology Department at the Armed Forces Institute of Pathology (AFIP) in Washington DC.

Anne Martel, PhD, is a Professor in Medical Biophysics at the University of Toronto, a Senior Scientist and the Tory Family Chair in Oncology at Sunnybrook Research Institute and a Faculty Affiliate at the Vector Institute. Her research is focused on medical image and digital pathology analysis, particularly on applications of machine learning for segmentation, diagnosis, and prediction/prognosis. Dr Martel is currently a senior editor for the journal Medical Image Analysis and previously served as an Associate Editor for IEEE Transactions in Medical Imaging. In 2006 she co-founded Pathcore (Toronto, ON), a software company developing complete workflow solutions for digital pathology.

George J. Netto, MD, is the Simon Flexner Professor and Chair of Pathology and Laboratory Medicine, Perelman School of Medicine, University of Pennsylvania. His research is focused on urologic and molecular diagnostic pathology and has helped characterize the incidence and role of TMPRSS2-ERG fusion and PTEN loss as a prognostic biomarker in prostatic adenocarcinoma. He is credited with the discovery related to the high incidence of TERT gene promoter mutation in muscle-invasive bladder cancer and upper tract urothelial carcinoma. Dr. Netto is an Editor-in-Chief of Modern Pathology and a Standing Editor of the upcoming 6th edition of the World Health Organization (WHO) Blue Book series.

Miguel R. Ossandon, PhD, is a Program Director in the Diagnostic Biomarkers and Technology Branch, at the Cancer Diagnosis Program, DCTD, NCI. He manages the NCI grant portfolio related to computational modeling and machine learning approaches for cancer diagnosis and digital image processing. He has a dual background in clinical laboratory and computer science. Prior to joining CDP, he worked in cancer research at the Lombardi Cancer Center, Georgetown University, where he developed interest in computer science. Miguel received his master’s degree at George Washington University and Ph.D. in computer science at the University of Maryland Baltimore County.

Liron Pantanowitz, MD, PhD, MHA, is the Chair and Professor of Pathology at the University of Pittsburgh. He is a hematopathologist, cytologist and is also board-certified by the American Board of Pathology in clinical informatics. Dr. Pantanowitz is an Editor-in-Chief of the Journal of Pathology Informatics. He is the president of the Digital Pathology Association, president of the American Society of Cytopathology, and a past president and current council member of the Association of Pathology Informatics. He is an expert in the field of pathology informatics and cytopathology. His research interests include digital pathology and artificial intelligence, as well as non-gynecologic cytopathology.

Anil Parwani, MD, PhD, is a Professor of Pathology and the Vice Chair and Director of Anatomic Pathology at the Ohio State University. He is also the Director of Pathology Informatics and the Director of Digital Pathology. His research is focused on diagnostic and prognostic markers in bladder and prostate cancer, and renal cell carcinoma. He has expertise in surgical pathology, viral vaccines and pathology informatics including biobanking, whole slide imaging, digital imaging, telepathology, image analysis, AI, and lab automation. Dr. Parwani is the Editor-in-chief of Diagnostic Pathology and Journal of Pathology Informatics.

Nilsa C. Ramirez, MD, FCAP, is a pathologist and the Director of the Biopathology Center (BPC) at the Nationwide Children’s Hospital and a professor of clinical pathology at the Ohio State University College of Medicine, Columbus, OH. Her expertise includes biobanking in the context of adult and pediatric cancer clinical trials. At the BPC she oversees the NCTN biobanking efforts of the Children’s Oncology Group, SWOG, and NRG Oncology-Columbus. She is a PI of the Pediatric Division of the Cooperative Human Tissue Network and a PI of the EET Biobank. Dr. Ramirez is a member of the Biorepository Accreditation Program (BAP) Committee at the College of American Pathologists (CAP) and the CAP BAP National Commissioner.

Dr. Joel Saltz is the Cherith Professor and Founding Chair of the Department of Biomedical Informatics at Stony Brook University, School of Medicine. His research focuses on digital pathology, artificial intelligence, and imaging biomarkers. Dr. Saltz received his MD and Computer Science PhD from the Medical Scientist Training Program at Duke University.

Sandro Santagata, MD, PhD, is a clinician-scientist practicing diagnostic pathology and leading a research team at Brigham and Women’s Hospital and Harvard Medical School. He focuses on developing and implementing new technologies and computational approaches for multiplexed tissue imaging of cancer resection specimens. These efforts are advancing the understanding of the biological properties and interactions of tumor and immune cells within the tumor microenvironment and identifying spatial biomarkers to improve cancer diagnosis and to tailor individual therapies to improve patient outcomes.

Lalitha K. Shankar, MD, PhD, is the Chief of the Clinical Trials Branch at the Cancer Imaging Program (CIP), DCTD, NCI. Her work involves establishment of and monitoring of clinical trials to evaluate imaging tracers and techniques, which aim to improve the prevention, diagnosis, and treatment of cancer. She provides imaging expertise for trials of cancer diagnostics and therapeutics sponsored by NCI in the NCTN and NCORP. Prior to joining the NCI, she was a faculty member in the Department of Radiology at Georgetown University and at the Lombardi Cancer Center and worked in the Division of Nuclear Medicine at Washington Hospital Center.

Amber Simpson, PhD, is the Canada Research Chair in Biomedical Computing and Informatics, and Associate Professor in the Department of Biomedical and Molecular Sciences and in the School of Computing at Queen’s University. She is an Affiliate of the Vector Institute for AI as well as a Senior Investigator at the Canadian Cancer Trials Group (CCTG). Dr. Simpson is the Director of the Centre for Health Innovation, a joint venture with Kingston Health Sciences Centre and Queen’s. She specializes in biomedical data science, focusing on developing novel computational strategies for improving human health.

Shakeel Virk, BS, is a Manager for the Canadian Cancer Trials Group (CCTG) Tumour Tissue Data Repository (TTDR). He also serves as a Director of Operations for the Queen’s Laboratory for Molecular Pathology (QLMP), Department of Pathology and Molecular Medicine at Queen’s University. He managed tissue banking activities for over 150 TTDR clinical trials. His lab provides immunohistochemistry, immunofluorescence, TMA construction, digital pathology, and DNA/RNA isolation services. Shakeel has a strong interest in digital pathology and image analysis and has set up the Queen’s Digital Pathology platform for scanning, sharing, annotating, and analyzing high resolution pathology slide images.

Kenneth C. Wang, MD, PhD, is a staff radiologist and MRI section chief at the Baltimore VA Medical Center, and Adjunct Associate Professor at the University of Maryland School of Medicine. He completed residency and fellowship training at the Johns Hopkins Hospital, and a fellowship in imaging informatics at the University of Maryland. He has worked on standards development with RSNA and LOINC, and his research interests include ontologies, interoperability, image segmentation, MR neurography, 3D printing, liver imaging, and shoulder surgery.

Mark Watson, MD, PhD, is the Margaret G. Smith Professor and Vice Chair of Faculty Development in the Department of Pathology and Immunology at Washington University School of Medicine. For the past 25 years, he has directed institutional and national biobanking efforts, emphasizing the use of informatics to develop robust solutions for managing complex biorepository operations. He is the Director of the Siteman Cancer Center Biorepository and of the NCI’s Alliance Group Biorepositories and Biospecimen Resource. His own research is focused on the identification and validation of genomic biomarkers to predict and mitigate metastasis in breast and lung cancer patients using spatial profiling, AI-based image analysis, and liquid biopsy analytes.

George Yousef, MD, PhD, is a Program Medical Director of Laboratory Medicine at University Health Network in Toronto, Canada, a Professor and Vice-Chair of the Department of Laboratory Medicine at the University of Toronto, and the Head of Informatics and Digital Resources at the Canadian Association of Pathologists. He is the Vice President of the International Society of Enzymology, the editor-in-Chief of the Canadian Journal of Pathology, and author of a book on “Molecular Pathology in Cancer”. His research laboratory is one of the leading laboratories focusing on the cancer biomarkers in renal cell carcinoma and prostate cancer.

Kun-Hsing "Kun" Yu, MD, PhD, is an Assistant Professor in the Department of Biomedical Informatics at Harvard Medical School. He developed the first fully automated artificial intelligence (AI) algorithm to extract thousands of features from whole-slide histopathology images, discovered the molecular mechanisms underpinning the microscopic phenotypes of tumor cells, and successfully identified previously unknown cellular morphologies associated with patient prognosis. His lab integrates cancer patients' multi-omics (genomics, epigenomics, transcriptomics, and proteomics) profiles with quantitative histopathology patterns to predict their clinical phenotypes.

George Zaki, PhD, is a director of applied scientific computing with Frederick National Laboratory for Cancer Research (FNLCR). He supports investigators across the National Cancer Institute in their scientific computing pipeline, data science, machine learning, and accelerated computing workflow development and sharing. George holds a Ph.D. degree in computer engineering from the University of Maryland.

Mark Zarella, PhD, is a scientist focusing on imaging and quantitative approaches in the Division of Computational Pathology and AI at the Mayo Clinic. Since 2012, he was on faculty in pathology, first at Drexel University College of Medicine as the Technical Director of Pathology Informatics, then as the Director of Digital Pathology at Johns Hopkins University. He joined the Mayo Clinic in 2022 where he helps to transition the department to an AI- and digital-enabled practice. Dr. Zarella also serves on the board of directors for the Digital Pathology Association and on the Digital and Computational Pathology Committee for the College of American Pathologists.