In the lab · three builds
Machine learning for the GI suite
Retained food on EGD. Computer vision that flags retained food during upper endoscopy, so the call gets made consistently, not by mood.
A microbiome screen for IBD. Machine learning over stool metagenomes to detect inflammatory bowel disease. It holds up on cohorts it never saw.
GI trial simulator. An open R package that re-runs the sample-size math behind published GI trials and reports which numbers survive.
A fourth bench is under a sheet.
github.com/tarek-arabi/gi-trial-simulator
In the lab · computer vision
Retained food on EGD
A model that flags retained food during upper endoscopy. Gastric contents change the read, the aspiration risk, and the repeat-scope decision, so the call should be made consistently, not by mood.
In the lab · machine learning
A microbiome screen for IBD
Machine learning over stool metagenomes to detect inflammatory bowel disease. Trained across public cohorts, and it holds up on cohorts it never saw.
In the lab · open source
GI trial simulator
An open R package that re-runs the sample-size math behind published GI trials and reports which numbers survive.
github.com/tarek-arabi/gi-trial-simulator