Artificial intelligence (AI) has been increasingly prevalent in a variety of fields, including medicine. Case Western Reserve University is on the forefront of this movement as it held an inaugural Artificial Intelligence in Oncology: Advancements and Policy Symposium event in Tinkham Veale University Center this past Thursday. The symposium was organized under the leadership of Anant Madabhushi, a faculty member at the CWRU School of Medicine, along with other members of the planning committee, and focused on AI in the study and treatment of cancer. The symposium brought together members of the scientific and medical community from CWRU, Cleveland Clinic, Harvard University, the University of Pennsylvania and more.
The symposium began with an address from President Barbara Snyder, followed by an address from Pamela Davis, dean of the CWRU School of Medicine.
Other prominent speakers included Hugo Aerts, associate professor at Harvard Medical School and director of the Program for Artificial Intelligence in Medicine at Brigham and Women’s Hospital, and Michael Feldman, professor of pathology and laboratory medicine at the Hospital of the University of Pennsylvania and the director of the Office of Pathology Informatics.
The keynote address was given by Sohrab Shah, professor at Weill Cornell Medical College and chief of computational oncology in the Department of Epidemiology & Biostatistics.
“Cancer is a disease of the genome,” said Shah in his opening remarks. He elaborated on that point by speaking about how artificial intelligence and mathematical modeling can be used to study the genetics of cancer. Shah explained that the modeling allele counts in breast cancer led to the discovery that new mutations had occurred after the initial tumor had formed and become notable, and thus explained why some treatment methods had varying degrees of success.
Furthermore, he explained how modeling allele populations over time can help show how a cancer is evolving, a process which can be expedited with the use of AI. Shah spoke about how ovarian cancer has massive genetic instability, citing a five-year survival rate and 80 percent recurrence as evidence. AI modeling could help study the aforementioned mutations and help combat them more effectively, potentially improving survival rates and reducing instances of relapse.
Shah’s findings were that the initial site contained a mixture of genetic populations but metastasis sites were composed of “unidirectionally genetic populations” (metastasis was caused by one type of mutated cell). Additionally, the clones that came to dominate metastasis sites were all present in the initial site.
Shah explained how, using a novel machine learning method, his lab was able to study gene expression in cancer cell populations. He expects multi-modal data integration to drive new research and lead to the development of new diagnostic tools, describing the implementation of a data model that automatically pulls data from over ten sources.
Madabhushi returned to the stage to speak about the radiomics program, which studies digital images of cells and automatically determines cell types, nuclei and more. The program can create arrow patterns and convert the resulting image into a thermal image, which can be used to study cancer. He explained how low-cost computational diagnostics can be particularly useful in improving healthcare in low-income countries.
A highlight among other speakers was Brandon Gallas, speaking about the U.S. Food and Drug Administration (FDA). He spoke about how a streamlined and proper filing for refined AI programs and studies can significantly expedite the process of rolling out programs rather than causing back-log at the FDA. Ensuring that concise explanations for mathematical calculations and having reproducible results will ensure quick processing.
Sharona Hoffman, a professor at both the School of Law and School of Medicine, spoke about how AI can help physicians treat patients in a more ethical way. However, there are concerns raised by such an advancement, such as risks to privacy and the threat of discrimination based off health information. She explained how the Americans with Disabilities Act defines disability, but doesn’t necessarily relate to future health problems; the Genetic Information Nondiscrimination Act prohibits discrimination based on genetic information, which includes predictive data, but only applies to employers and health insurers and only covers future genetic health problems. There is no law regarding future non-genetic health problems as it relates to discrimination.
Hoffman explained that her long-term predictions for AI in healthcare include improved detection and treatment of a variety of health problems, but that privacy and anti-discriminatory laws must keep up with advancements in AI to ensure ethical implementation. AI should also be made more accessible to all people, instead of being primarily for the wealthier members of society. Additionally, psychological care could be provided to those who receive impactful or potentially traumatizing information from genetics-based or AI programs.
A final panel was held and questions were asked about the importance of privacy in the advancing medical field, particularly as medical information is becoming increasingly accessible.
The symposium has been approved by President Snyder to become an annual event. Madabhushi hopes to add a greater variety of speakers and panelists in future AI in Oncology Symposiums.