OnDemand: AI for Obesity Research and Clinical Practice-AI-Driven Solutions in Health Research

Video Transcription

Video Summary

Dr. Satish Vishwanathan discussed how AI, machine learning, and deep learning are transforming healthcare through “pixels to predictions.” He focused on image analytics in cancer and digestive disease, while also touching on obesity research. He explained AI’s evolution from rule-based systems to statistical machine learning and modern deep learning models, then highlighted the growing use of AI in healthcare and cancer research.<br /><br />A major theme was precision medicine: using imaging and clinical data to predict outcomes and tailor treatments to individual patients. Examples included radiomics and pathomics, where features are extracted from MRI, CT, and digital pathology images to classify disease, predict therapy response, and assess risk. He showed studies using MRI scans in colorectal cancer and Crohn’s disease, and noted that combining imaging with clinical data improved performance.<br /><br />Vishwanathan emphasized two critical issues: interpretability and generalizability. He described how models can mistakenly learn irrelevant cues, like rulers in melanoma images or backgrounds in dog photos, instead of true disease signals. He also showed that models often fail when applied to new institutions unless sources of variation are carefully controlled. Finally, he discussed human-in-the-loop workflows and emerging prompt-based AI tools, stressing that AI is promising but must be transparent, reliable, and clinically useful.

Asset Subtitle

By Satish Viswanath, PhD

Keywords

artificial intelligence

machine learning

deep learning

precision medicine

radiomics

pathomics

healthcare analytics