Digital Pathology Implementation: Insights From Experts at DP&AI: USA

The conference commenced with two thought-provoking keynote addresses that set the tone for the entire event. This is part two of our report on the Digital Pathology & AI Congress: USA. For part one, please see links toward the bottom of the article.

Preparing for and Deploying All-Digital Workflow at Significant Scale: The University of Michigan Experience

Ulysses G. J. Balis and Mustafa Yousif of the University of Michigan shared their experience in preparing for and deploying an all-digital workflow at a large scale. They emphasized the importance of thorough planning and organizational changes when transitioning to an all-digital workflow for primary diagnosis in large-scale pathology practice. Their presentation highlighted the key aspects of image management system selection and deployment, as well as the integration of AI-based solutions.

Dr. Balis presented key statistics for Michigan Medicine Pathology, which currently hosts 165,000 surgical specimens and over 100,000 transfer and consult cases and has 10%-12% annual year-over-year volume growth.[1] With seven remote locations needing frozen sections and consultative support, the department is actively integrating recently acquired health systems.[1] Benefits of digital pathology implementation at the University of Michigan include faster turnaround times, increased diagnostic accuracy, the possibility of multidisciplinary tumor board meetings, and enhanced communication between pathologists and other members of the care team.

Dr. Balis also discussed the 5-year digital transformation goals of the department, which include plans to integrate digital pathology with molecular omics (e.g., spatial transcriptomics) to allow for the development of predictive models. He also discussed the department’s post-modern view of the full portfolio of possible platform services enabled by cross-enterprise integration. These services include remote consultations and support, real-time consultation, image-based practice tools, image-based query and analysis tools, educational tools for the general public and pathology departments, integrated diagnostic reporting, multi-axial drug discovery tools, reference image libraries, and ‘patients like me’ tools. The department also plans to integrate AI into the digital pathology workflow to screen for diagnostic patterns, predict treatment responses, risk-stratify patients, assess billing completeness, and determine thresholds for seeking consultation. During the process, user satisfaction will be continuously monitored, and the tools and their implementation will be adjusted according to the feedback.

“Over time, the nice to have becomes a must-have,” noted Dr. Balis when discussing their plans of using Innovation Ambition Matrix to integrate pathology with next-generation sequencing and other omics technologies.

Commenting on challenges faced when deploying digital pathology for primary diagnosis at the University of Michigan, Dr. Balis said, “Overcoming natural skepticism of rank-and-file pathologists that digital sign-out could be as effective as that provided by microscopy is a key cultural challenge.”

Cost is another critical barrier to the large-scale implementation of digital pathology. “We ultimately decided to utilize departmental reserves, as this project was felt to be existentially important for us,” noted Dr. Balis. He explained that the initial investment is less than 10 million, but costs quickly increase to 25-30 million with long-term storage.

Dr. Balis, Associate Chief Medical Information Officer and A. James French Professor of Pathology Informatics, University of Michigan.

Dr. Balis, Associate Chief Medical Information Officer and A. James French Professor of Pathology Informatics, University of Michigan.

Dr. Yousif shared strategies for effective management changes, site preparation, and cockpit design to ensure a smooth transition to digital workflows.[2] The department leveraged the existing infrastructure of Michigan Medicine and enterprise resources, highlighting that the digital pathology infrastructure can be siloed and fragmented.

A key challenge when a pathology lab is going digital is the number of slide scanners required. Dr. Yousif explained that they analyzed the scanning throughput for internal slide scanners, slide distribution, and the number of slides tested. Following this assessment, they determined that seven scanners would be ideal for the department.

Another important question is where the scanners should be placed in the lab. Based on their experience, they determined that the ideal scenario for their lab was to have scanners in the center of the room. This placement allows scan technologists to pick slides directly from the stainer or coverslip, reduces the overall walk for multiple workflows, improves immunohistochemistry cutting, and involves less handling of the slides.

They also assessed different scan workflows and found that having one reading and sign-out room in the center of the lab worked well. Factors for ergonomic workflow to consider when establishing a reading and sign-out room include lighting, furniture selection, temperature control, noise reduction, technology integration, organization, and accessibility.

Dr. Yousif also discussed the crucial role of case management workflow and phased service rollout in facilitating the smooth integration of digital workflows. He also touched on the importance of effective communication with stakeholders before project commencement and throughout the project, comprehensive training for faculty and staff transitioning to digital pathology, and robust quality assurance measures to ensure the accuracy and reliability of digital pathology processes.

Dr. Yousif, Assistant Professor and Director of Digital Pathology, University of Michigan.

Dr. Yousif, Assistant Professor and Director of Digital Pathology, University of Michigan.

Transforming Pathology Through Digital Innovation: The Mayo Clinic Experience

Jason Hipp of the Mayo Clinic delivered a keynote talk on transforming pathology through digital innovation, sharing his insights from his experience at the Mayo Clinic.[3] He began his talk by highlighting the transformative potential of computational pathology and AI, which allows pathologists to move from qualitative to quantitative data and perform assessments that are impossible with standard pathology.

Digital pathology relies on the convergence of laboratory medicine, pathology, technology, and data sciences. Although digital pathology emerged over 20 years ago, less than 10% of pathology departments in the US have adopted digital pathology solutions. However, recent advances in cloud technologies, big data, and AI have opened new avenues and boosted the implementation of digital technologies in pathology departments and large clinics, including Mayo Clinic.

Dr. Hipp provided insights into how the Mayo Clinic leverages computational pathology and AI to advance clinical practice, patient care, and biopharmaceutical research, showcasing practical applications and use cases. Approximately 190,000 slides are scanned monthly across all Mayo Clinic sites. The clinic has 37 Leica GT450 scanners across all sites, opening up 38 enterprise-wide positions to support digital pathology.

Mayo Clinic is implementing a tissue registry program to digitize their archives containing over 25 million tissue glass slides. They plan to digitize 12 million slides by the fall of 2024. The digitized data will be stored in a cloud-based image repository to support clinical, research, and educational use. Moreover, 3 million historical paper pathology reports will be digitized to support longitudinal pathology studies. The clinic has also developed AI systems for in-line automated quality assessment and corrections for archival scanning. Dr. Hipp explained that integrating AI models into their scanners allows them to perform edge computing, analyze images in real time, and obtain a 3D spatial representation of the tumor microenvironment through volumetric scanning.

Dr. Hipp also provided insights into the exciting future possibilities of computational pathology and AI in revolutionizing patient care. In collaboration with Aiforia, Mayo Clinic developed AI SANBOX@MAYO, an AI research pipeline that allows pathologists to translate projects from ideas to practice. The research pipeline fosters innovation and upskills in pathologists by allowing pathologists who are not experienced with AI to leverage AI tools.

The clinic has also developed a multimodal platform that allows pathologists to collect structured or unstructured biomedical data, fully digitized EMR data, digitized pathology slides (12,000,000 by fall 2024), a collection of 9,900,000 ECG tests, a diverse CAP/CLIA diagnostic test menu (3,800), and radiology images (241,000,000 CT scans, 146,000,000 MRI scans, and 2,500,000 PET scans). In collaboration with technology partners, the clinic also aims to create next-generation spatial biomarker data.

By leveraging advanced digital technologies and cloud systems, Mayo Clinic has been able to enhance diagnostic accuracy and streamline workflow processes, ultimately improving patient care outcomes. Dr. Hipp highlighted the potential of these technologies to revolutionize patient care by enabling more personalized and precise treatments and accelerating drug development processes through advanced computational capabilities. He underscored that a fully immersive visual experience combined with touch, vision, and sound enabled by spatial computers can enhance the clinical implementation of digital technologies in an ecosystem in which pathologists can work seamlessly with computers to fully realize the benefits of practicing digital pathology.

Dr. Hipp, Chief Digital Innovation Officer at Mayo Collaborative Services, Mayo Clinic.

Dr. Hipp, Chief Digital Innovation Officer at Mayo Collaborative Services, Mayo Clinic.

Links To Other Parts Of The Series

Part 1: Highlights from the 10th Digital Pathology & AI Congress: USA

References

[1] Ulysses G. J. Balis. Preparing for and deploying all-digital workflow at significant scale: the University of Michigan experience. Presented at the 10th Digital Pathology & AI Congress: USA; May 7-8, 2024; San Diego, CA.

[2] Mustafa Yousif. Preparing for and deploying all-digital workflow at significant scale: the University of Michigan experience. Presented at the 10th Digital Pathology & AI Congress: USA; May 7-8, 2024; San Diego, CA.

[3] Jason Hipp. Transforming pathology through digital innovation: The Mayo Clinic experience. Presented at the 10th Digital Pathology & AI Congress: USA; May 7-8, 2024; San Diego, CA.

Share This Post

Leave a Reply