Arya
summary
A Bayesian generative disease‑progression model implemented in Stan uses a latent severity axis (age on x), four‑parameter logistic biomarker curves, I-splines and hierarchical priors to borrow strength across patients and produce full uncertainty over trajectories. The work claims concrete numbers: 11% misdiagnosed and false positives costing $14k/year, and shows how to compute drug benefit probabilities by sampling paired trajectories. Stan posterior predictive checks sometimes match real biomarkers and sometimes reveal mismatched patient‑specific variance, guiding model refinement.
AI-written summary · about
chapters (10)
- 0:00Alzheimer's diagnosis research
- 2:12Accurate diagnosis importance
- 4:01Manual diagnosis limits
- 5:48Bayesian generative modeling
- 8:00Generative vs classification
- 10:12Disease progression nonlinearity
- 11:41Biomarker variance insights
- 12:47Uncertainty in treatment
- 14:54Multi-level modeling benefits
- 15:57Q&A
related talks
26mMachine Learning for Personalised Health
28mStan: why does it exist? when is it useful? why do I use it?
01h 24mBayesianism and Survival Analysis
34mFast, beautiful and easy Bayesian modelling: Can you have it all?
56mIntroduction to Cancer Bioinformatics III: Cancer Progression Networks
21mHazards on the Causal Path
31mCounterfactual Analysis with Bayesian Models
31mBayesian Risk Analysis For Large Multi-Modal Data