Patient Specific Digital Twins Beyond Average Models Tom Yankeelov

Описание: Tom Yankeelov (UT Austin) presents a strategy for building clinically actionable digital twins for cancer patients using mechanism-based mathematical models calibrated entirely with patient-specific imaging data.
He begins by arguing that traditional clinical trials optimize treatments for the “average patient,” a concept that does not meaningfully exist once you consider deeper biological heterogeneity. Relying only on population-level statistical inference risks missing the nuances that determine an individual patient’s response.
To capture patient-specific behavior, his team uses 4D MRI data (3D space + time) that provide anatomical structure, cellularity, and perfusion. These measurements feed into a partial differential equation model describing tumor cell movement, proliferation, therapy response, and drug delivery. Crucially, the model is calibrated only to data from that individual, not population averages.
The model demonstrates strong predictive accuracy across multiple cancers (breast, prostate, cervical, brain) without retraining each time—a major distinction from many AI approaches that require disease-specific retraining.
Using these calibrated models, the team constructs digital twins by simulating many possible treatment regimens—including different dose schedules—to identify the optimal strategy for each patient. Their simulations show that standard regimens are often far from optimal for many individuals. They demonstrate applications to chemotherapy, targeted therapy, and radiation therapy, including dose-painting based on predicted proliferation.
The team also validates digital-twin-optimized regimens in mouse studies, showing significantly better responses with less chemotherapy through optimal control theory.
The talk concludes with key implementation challenges: workflow logistics, physician/patient acceptance, data transfer, integration with radiology/pathology workflows, and defining the first clinical use cases.
If you found this video useful, please check out our other videos on computational modeling, infection and immunology:
Please consider joining our IMAG/MSM WG on Multiscale Modeling and Viral Pandemics: https://www.imagwiki.nibib.nih.gov/co...
Please also consider joining the Global Alliance for Immune Prediction and Intervention: http://glimprint.org/
#DigitalTwins #PrecisionOncology #CancerResearch #MedicalModeling
#PersonalizedMedicine #MathematicalOncology #MRIImaging
#CancerTreatment #BiomedicalEngineering #HealthcareAI