Abzal Khairullauly CT scan annotation tool for stroke detection
Frontend Developer · Lean Solutions · Pilot (frozen)
Angular TypeScript Canvas API
Problem
Radiologists needed a browser tool to label brain CT scans, producing training data for a stroke-detection ML model.
What I did
Canvas-based image viewer: zoom, pan, fit-to-frame, and window/level contrast control by mouse drag Diagnostic measurements: Hounsfield-unit probe, circular ROI with mean HU, ruler in mm Segmentation editor: brush, eraser, closed and open polygons, scissors (clip to region), threshold by pixel range, mask opacity Model-assisted labeling: load the ML model's prediction as a starting point, then correct it by hand Classification mode: multi-class and multi-label Editor UX: keyboard shortcut for every tool, multi-step undo, per-class visibility toggles Workflow and roles: In progress → Finished → Edit states; labelers edit while managers and admins get read-only tools; unsaved-changes guard
Results
Canvas-based image viewer: zoom, pan, fit-to-frame, and window/level contrast control by mouse drag Segmentation editor: brush, eraser, closed and open polygons, scissors (clip to region), threshold by pixel range, mask opacity Model-assisted labeling: load the ML model's prediction as a starting point, then correct it by hand
Media
