
Why Medical AI is Solving the Wrong Problem (and How to Fix It) | Smart Medicine Ep. 1 đ©șđ»
AI is the hottest topic in medicine, but why are so many clinical AI tools failing to deliver a real ROI?In this premier episode of the Smart Medicine Podcast, host Dr. Ali Fatemi (clinical medical physicist) sits down with Dr. Khan Siddiquiâa rare "quadruple threat": a practicing radiologist, academic researcher, software engineer, and deep-tech entrepreneur.Dr. Siddiqui has spent his career at the absolute cutting edge of technology. He learned to code in the 6th grade on a PDP-11 gamma camera system in Pakistan, built epidemiological linear regression models for the CDC in 1993, and went on to work at Microsoft, where his team's machine learning work laid the foundation for the Xbox Kinect, Azure ML, and the revolutionary ResNet neural network architecture. Later, he served as Chief Medical Officer and Chief Strategy Officer at Hyperfine, co-inventing the world's first portable bedside MRI. Today, he is the founder and CEO of Hopper, building multimodal foundation models to streamline clinical workflows.In this episode, Dr. Siddiqui breaks down why improving lesion detection is NOT the key to clinical AI success, explains why AI will never replace radiologists, and shares how we can build AI that actually saves clinicians from crushing burnout.â±ïž TIMESTAMPS TO WATCH:00:00 â Introduction & Dr. Siddiqui's remarkable background.02:15 â Coding in the 6th grade & building linear regression models for the CDC in 1993.06:30 â Choosing radiology and pioneering the world's first imaging informatics fellowship.12:00 â The Microsoft years: Random forests, Xbox Kinect, Azure ML, and the birth of ResNet.16:15 â Redefining medical imaging: The story behind Hyperfine's portable bedside MRI.21:30 â The Windows Kernel Story: "Treating a complex engineering bug like a difficult patient".24:00 â Why clinical AI tools fail: The 250-millisecond detection myth vs. workflow reality.33:00 â What is driving the crushing rise in global imaging volume?.40:30 â Demystifying Multimodal AI & vector representations in medical imaging.48:00 â Hopper & Presto: Creating an "ambient scribe" that integrates directly into existing templates.54:30 â Will AI replace radiologists? Jevons Paradox, horse carriages, and the autopilot analogy.59:00 â Dr. Siddiqui's personal "Why": Helping his overworked radiologist wife reclaim her evenings.1:02:15 â Personalized medicine, AlphaFold, and losing 130 lbs using data and LLMs.đĄ KEY TAKEAWAY FROM THE EPISODE: "You always innovate at the edge of your experience." To solve healthcare's biggest challenges, we must stop building isolated AI "point solutions" and start building generalized models that adapt directly to clinicians' existing workflows.Subscribe for more deep dives into the future of smart medicine and medical technology! đ
Fler avsnitt frÄn "Smart Medicine Podcast with Dr. Ali Fatemi"



Missa inte ett avsnitt av âSmart Medicine Podcast with Dr. Ali Fatemiâ och prenumerera pĂ„ det i GetPodcast-appen.








