Personalized Medicine, Diabetes, and AI with Dimitris Bertsimas

In this MIT Learn Beyond Biology episode, MIT professor Dimitris Bertsimas joins Mary Ellen Wiltrout to share how advances in data and AI are helping healthcare move beyond one-size-fits-all treatment. Dimitris discusses how operations research, AI, and patient data could help make medicine more personal with a focus on diabetes care. Dimitris shares how family experiences shaped his healthcare research, including work using electronic medical records to help personalize diabetes treatment. He also explains how genomic data could support earlier intervention, why sharing anonymized health data matters, and how AI might help clinicians make decisions without replacing their judgment. The conversation ranges from optimizing Panama Canal operations to the future of personalized medicine and lifelong learning. Dimitris Bertsimas is the Vice Provost for Open Learning, the Associate Dean of Online Education and Artificial Intelligence, Sloan, and the Boeing Professor of Operations Research.Key TopicsWhat operations research is: Dimitris defines operations research as a scientific approach to designing, improving, and optimizing systems, with applications ranging from transportation to healthcare.A real-world optimization example: He recounts how modeling helped improve vessel scheduling and throughput at the Panama Canal.Why healthcare became a focus: The experiences of losing his father to gastric cancer and his mother to diabetes deepened his commitment to healthcare research.Personalizing diabetes care: His team worked with long-term electronic medical records to study how treatment combinations could be tailored to individual patients.The role of genomics: Dimitris discusses research into beta cell decline and the potential to identify diabetes risk earlier.His view of AI: He distinguishes analytics using structured data from AI approaches that can work with unstructured data, such as text and images.Why data sharing matters: He explains how anonymized patient data can support research while removing identifying details.AI in emergency departments: His group is piloting systems that combine triage notes and patient records to provide clinicians with a differential diagnosis and test suggestions.AI as clinical support: Dimitris emphasizes that AI should advise and assist; physicians remain responsible for decisions.Lifelong learning: He reflects on adapting to new fields and shares how MIT Open Learning serves learners at different stages of life.Timestamps00:00 Why medicine is still largely not personalized01:13 What operations research means02:15 Using optimization to improve Panama Canal operations04:10 How family experiences led Dimitris toward healthcare research05:53 Why diabetes became a personal and research priority06:29 Type 2 diabetes and approaches to managing the disease09:10 Studying long-term diabetes records10:08 Using patient history to guide treatment choices11:03 Early AI research in personalized diabetes care12:24 Studying beta cell decline with genomic research13:47 Dimitris’ distinction between analytics and AI15:47 What personalized or precision medicine means16:45 A personal story illustrating why cancer treatment can affect patients differently18:22 How continuous glucose monitoring informs Dimitris’ own diabetes management20:44 Why sharing anonymized health data can advance research23:09 How anonymization works24:47 Generative AI pilots for emergency department diagnosis support26:15 Using AI to suggest diagnoses and tests for physician review27:13 Reported effects on emergency department stays and insurance test rejections28:44 Why AI should advise rather than make medical decisions30:38 Dimitris’ five-year outlook for personalized medicine33:03 Adaptiveness, impact, and continuous learning35:29 Applying ideas through entrepreneurship and iteration36:51 MIT Open Learning and lifelong learning resources