Generative AI for Medical Imaging
My thesis work on synthetic chest X-rays: when they help a tuberculosis classifier, is it because they add new variety, or just because there are more images? Part of it became a short paper at Canadian AI 2026.
Connor Pink
I'm doing an M.Sc. in Computer Science at the University of Guelph, in Dr. Sykes' AI and Healthcare Lab. Most of my research is on chest X-rays: segmenting lungs, generating synthetic scans, and testing whether those synthetic scans actually help a tuberculosis classifier.
Before that I did two co-op terms building internal software, at Employment and Social Development Canada and at Peterborough Regional Health Centre.
Now
Do synthetic X-rays help, or is it just more data?
My thesis looks at training a TB classifier on mixes of real and generated chest X-rays (from a WGAN-GP and a latent diffusion model), to separate two effects: new variety in the data versus simply having more images. Part of it became a short paper at Canadian AI 2026. The diffusion images look better, but neither generator beat just duplicating real images to the same count. Read the paper or the project notes.
Research
CXR Segmentation and TB Imaging Pipeline
A chest X-ray research pipeline combining tuberculosis classification, lung segmentation, and diffusion generation, with a current focus on segmentation model comparison and efficiency tradeoffs.
ML-Based Bug Detection System
An applied research project for Nevvon that modeled application freezes through session analysis, stuckness heuristics, and predictive factor discovery.
Projects
Stencil-AI
A dual-path stencil generation system that combines Stable Diffusion and classical computer vision to create clean print-ready stencil artwork.
React Movie Watchlist
A full-stack movie watchlist application with account-based persistence, TMDB integration, editable saved entries, and containerized backend deployment.
sendrecv
A Go CLI for repeat SSH-based file transfer between known devices, with host presets, archive logic, remote checks, and Yazi integration.
Raycast Overseerr Requesterr
A Raycast extension for searching media through Overseerr, checking availability, and sending requests into a Radarr / Sonarr workflow.