Home▸Bioinformatics▸BraTS Brain Tumor Segmentation — Dice Score Explorer

🧠 BraTS Brain Tumor Segmentation — Dice Score Explorer

Adjust a simulated AI tumor-segmentation region against a fixed ground-truth brain tumor outline and watch the Dice similarity coefficient and a Hausdorff-distance-style boundary mismatch update live, the same metrics used to score real BraTS brain tumor segmentation challenge submissions.

Bioinformatics2DModerate60 FPS

This simulation models how automated brain-tumor segmentation AI is scored against a human-drawn "ground truth" region, the core evaluation method used in the real BraTS (Brain Tumor Segmentation) challenge. A fixed irregular blob represents a radiologist-annotated tumor outline; a second, adjustable blob represents an AI model's predicted segmentation, scored live with the Dice similarity coefficient and an illustrative boundary-mismatch indicator.

🔬 What It Demonstrates

🎮 How to Use

Drag the horizontal and vertical offset sliders to shift the predicted (cyan, dashed) segmentation away from the fixed ground-truth (red) tumor outline, adjust the size slider to make it too small or too large, and increase jaggedness to make its edge noisier and less anatomically plausible. Watch the Dice score, the overlap quality band, and the boundary-mismatch indicator update live as the two outlines drift apart or converge.

💡 Did You Know?

Did you know that in the real BraTS challenge, teams are scored on three overlapping tumor sub-regions at once, whole tumor, tumor core, and enhancing tumor, each with its own Dice score, because a model can localize a tumor well overall while still missing its most clinically important enhancing core?

⚙ Under the hood

BraTSbrain tumor segmentationDice scoremedical image segmentationHausdorff distanceMRIdeep learning evaluationcomputational biology

2D · HTML5 Canvas 2D · 60 FPS target · runs fully client-side, no install