Miimansa recently hosted a webinar on AI for Advancing Evidence-Based Practice in Oncology, featuring insightful discussions with experts Dr. Kshitij Joshi (Co-founder, MOC – a leading cancer care chain in Mumbai), Dr. Gayatri Tahiliani (DNB Surgical Oncology), Dr. Abhishek Chauhan (DM Critical Care), and Dr. Seema Jagiasi (Medical Oncologist, MOC Kemps Corner).
The session explored how AI can support clinical decision-making, enhance data-driven oncology practices, and bridge the gap between guidelines and real-world care.The discussion highlighted the importance of structured data capture—whether from PET scans, EMRs, or molecular reports—to improve AI-driven insights and optimize patient outcomes.
AI-Assisted but Human-Led
Recent studies on the performance on AI models on tasks such as guidelines retrieval and clinical chart abstraction have reported encouraging results but these appear to be far from “clinincal-grade”. Furthermore, clinicians rely on guidelines, clinical experience, and real-world constraints like affordability and drug availability—crucial factors for clinical decision making that appear to be challenging for the current generation of AI models
Difficult to Treat
Poor treatment response, short-lived remissions and relapses drive patients towards indication for advanced treatment lines, making clinical decisions increasingly multi-factorial and complex. “Patchy” information availability is a key challenge in such patients. Even though a number of molecular markers are known candidates for personalized therapy selection and prognosis, information regarding their status may be hard to come by.
Decision making for such difficult cases can be exceptionally challenging. Solutions that can efficiently search a growing body of knowledge from clinical studies, case reports, practice guidelines, including real-world sources such as historical medical records, claims and patients’ blogs that can augment the imperfect view of a patient’s condition are needed.
Recommendations backed by Evidence
Explainability of AI generated recommendations is important for their inclusion in the treatment decision workflow. Practice guidelines related to a given condition can seldom misalign. Support for a particular recommendation through clinical history, external knowledge sources and secondary analyses builds trust in the veracity of AI generated recommendations and is essential for their adoption in clinical practice
AI for Oncology
Domain-adapted AI systems are increasingly demonstrating their superiority over larger, more general foundational models, particularly in specialized fields.. This is because these models are trained on a vast corpus of general data, which may not adequately represent the intricacies and specific patterns within a particular area, such as cancer biology.
Our understanding of neoplastic disease pathways is constantly evolving, resulting in an exponential expansion of available knowledge sources. This dynamic landscape necessitates AI models that are tuned on content from the latest research and clinical practices. For AI to truly transform oncology, better data organization and India-specific solutions are essential. A future where oncologists can instantly access patient summaries, treatment guidelines, and relevant molecular insights seamlessly appears within reach. Building such a solution requires collaboration between clinicians, researchers, and AI developers—a journey Miimansa is committed to taking on.
Comments