Artificial intelligence is rapidly transforming precision medicine by uncovering hidden genetic disease drivers, enabling continuous patient monitoring through wearable technology, and improving diagnostic accuracy in medical imaging. At the 5th Binaytara Precision Oncology Summit, experts highlighted how AI-driven tools are reshaping cancer research, personalized treatment strategies, and preventive care across oncology and chronic disease management.

This article summarizes the transcript from Session 1 of the 5th Binaytara Precision Oncology Summit. It has not been reviewed by the speaker and may contain errors. Chaired and moderated by Sylvia K. Plevritis, PhD (Stanford University), the session featured a presentation by Haruka Itakura, MD, PhD (Stanford Medicine) on predicting cancer outcomes with AI.

Dr. Haruka Itakura

The Big Data Challenge in Modern Medicine

Working with large-scale biological datasets has traditionally required interpretation through complex statistical analyses. However, artificial intelligence is fundamentally changing how we approach medical data analysis. While AI applications in patient management have received significant attention, the technology offers equally transformative potential in research and diagnostics.

Discovering Genetic Disease Causes

Beyond Known Disease Genes

The American College of Medical Genetics and Genomics (ACMG) maintains a list of genes known to cause genetic diseases. However, current knowledge covers only about 30 percent of heritable diseases. What accounts for the remaining 70 percent?

Traditional approaches like polygenic risk scores attempt to address this gap but often perform poorly. The challenge lies in how genetic variants interact—genes work together in nonlinear, complex ways that simple models cannot capture.

AI-Powered Gene Discovery

New AI approaches can identify nonlinear interactions between groups of genes, revealing which combinations appear in disease cases. This methodology has successfully identified new genetic contributions to conditions like thoracic aortic aneurysm, a highly heritable condition (70 percent heritability) that previously lacked a complete genetic explanation.

These methods have expanded beyond individual conditions, with applications in cases where genetic signals follow irregular patterns. For diseases like [unclear], known to have a high heritability but an incompletely understood genetic basis, AI methods have identified 10 to 100 times more contributing genes, at least quadrupling our understanding of genetic causes.

Wearable Technology and Continuous Monitoring

From COVID Testing to Comprehensive Health Tracking

The development of systems for monitoring COVID-19 and other infectious diseases through wearable devices represents just the beginning. AI enables the extraction of far more information from simple sensors than previously imagined.

Smartwatch data alone can generate predictive models for:

• Red blood cell counts
• Blood glucose levels
• Multiple other clinical measurements

While not as precise as dedicated laboratory tests, these models are accurate enough for meaningful health tracking. Measurements once requiring laboratory visits, glucose monitoring, and physiological parameters can now be estimated continuously from wearable sensors.

Electrodermal Activity and Beyond

Wearables measure parameters like electrodermal activity (skin moisture changes), which provide additional health information when processed through AI algorithms. These continuous, passive measurements create comprehensive health profiles without requiring active testing.

Revolutionizing Diabetes Management

Understanding Type 2 Diabetes Heterogeneity

Research into type 2 diabetes has revealed that it is not a single condition but rather encompasses many subtypes, potentially ranging from type 2A through 2Z. Different subtypes show distinct patterns in hemoglobin A1C levels and other markers.

Continuous Glucose Monitoring Insights

Continuous glucose monitors provide revolutionary insights into individual metabolic responses. Studies show that approximately 90 percent of people demonstrate some degree of pre-diabetic glucose patterns. Wearing these monitors fundamentally changes eating behavior; individuals virtually never eat the same way once they see their glucose responses in real-time.

Different individuals show dramatically different glucose patterns, even for the same foods. Machine learning can predict these patterns and classify diabetes subtypes just from the shape of glucose curves. The type of diabetes—whether primarily insulin resistance or beta cell dysfunction—can be determined from monitoring data, which then informs treatment selection.

Personalized Nutrition and Lifestyle

AI enables matching diabetes subtypes with optimal foods and eating patterns. Personalized recommendations consider not just what to eat but when to eat it. By breaking down foods into their molecular components and using databases of nutritional information, AI can create truly individualized dietary plans.

This personalization is crucial because even when people want to follow recommendations, they often struggle if the advice doesn't account for their preferences and lifestyle. AI-powered programs can create sustainable, personalized plans that people will actually follow.

Medical Imaging Revolution

Beyond Traditional Analysis

Medical imaging represents one of AI's largest application areas, with approximately 17 different AI tools now available for various imaging modalities. The technology extends beyond obvious applications to identify associations not previously recognized.

Diagnostic Performance

For conditions like glioma diagnosis, the best traditional studies achieved an accuracy of around 83-84 percent. Image analysis using AI achieves similar or better performance. For many applications, AI systems now match or exceed human expert performance.

Current approaches analyze phenotypes in unprecedented detail. The field continues to advance, with some systems analyzing 12 or more distinct features simultaneously.

Preventive Medicine Applications

Whole Body MRI Screening

Whole-body MRI screening remains controversial, primarily because critics argue that it identifies issues that will never cause problems. However, proponents note that these scans do detect significant pathology, including cancers. The key question is not whether abnormalities exist but whether they are growing.

AI interpretation of these scans addresses some concerns by more accurately distinguishing benign findings from those requiring attention. Many people are now using these screening services, with AI providing interpretation through automated analysis systems. In the short term, this technology will likely become standard for screening in many contexts.

Future Directions

The integration of AI across precision medicine continues to accelerate:

• Genetic discovery expanding to cover more heritable diseases
• Wearable technology providing continuous health monitoring
• Diabetes management becoming truly personalized
• Imaging interpretation becoming more accurate and accessible
• Preventive screening becoming more sophisticated

These advances represent the current state of AI in precision medicine, with applications expanding rapidly. The technology is not replacing clinical judgment but rather augmenting it, providing insights and capabilities that enhance patient care across multiple domains.

Looking Forward

AI has moved from experimental technology to a deeply embedded tool in precision medicine. The molecular signatures being discovered, the continuous monitoring capabilities being developed, and the personalized treatment approaches being validated all rely heavily on AI's ability to find patterns in complex biological data.

The goal is not just more data but better insights that lead to improved patient outcomes. As these technologies mature, the integration of genetic information, continuous monitoring, and AI-powered interpretation will become standard components of medical care.