← Back to work
Health & public health · Education & research
CELL_LOG
Make medical-image model output inspectable through explanations and history.
- Status
- Academic research
- My role
- Independent research, AI, API, and interface implementation
- Timeframe
- May - June 2026
- Core technology
- FastAPI · TensorFlow · OpenCV · React
01Separate breast and lung pipelines
02Grad-CAM explanations
03Academic exhibition selection
01 / Context
The problem
Image classifiers are difficult to inspect when results are presented without visual context or longitudinal history.
02 / Product
The implemented solution
CELL_LOG connects model output with heatmaps, scan history, aggregate views, and exportable research reports.
03 / System
Architecture, in plain language
Separate model services sit behind a tested FastAPI layer and a responsive research interface.
Interface→Application services→Data & integrations
04 / Judgment
Important decisions
- Separate model pipelines.
- Show explanations beside outputs.
- Use decision-support language throughout.
05 / Boundaries
Limitations and next work
The models are not clinically validated and must not be used for diagnosis or treatment decisions.
06 / Implementation
Focused technology
FastAPI · TensorFlow · OpenCV · React · SQLAlchemy · Docker