Data & AI · Computer Vision
Kidney Stone AI Thesis.
An undergraduate thesis exploring YOLO-based kidney-stone detection and bounding-box-based size estimation from ultrasound images.
01
Research focus
The thesis examines how object detection can identify suspected kidney stones in ultrasound imagery and support a measurement workflow.
02
Technical work
The project covers dataset preparation, image preprocessing, object detection, model evaluation, and web-based system integration.
- Prepare and structure ultrasound image data
- Evaluate detection behavior and measurement logic
- Translate experiments into a usable decision-support interface
03
Product perspective
Beyond model performance, the work asks how an AI experiment can become a legible system for human review without overstating what the model can decide.
04
Evaluation snapshot
The academy portfolio documents 3,006 training images, 971 validation images, and 1,023 test images, with a reported all-classes mAP@0.5 of 0.96. These are research-prototype results, not a clinical claim.