Enhancement: Real-time Segmentation Metrics (Dice/IoU)

[blayyyyyk]

Description

As a user, I want to see Dice and IoU metrics displayed in the browser so that I can quantitatively evaluate the model's performance during the live presentation.

Acceptance Criteria

• Load a ground-truth spleen mask alongside the base medical volume in the viewer.
• Write a client-side JavaScript utility to calculate the Dice Coefficient between the SAM 3 prediction array and the ground-truth array.
• Write a client-side JavaScript utility to calculate the Intersection over Union (IoU).
• Create a UI element (e.g., a floating HTML panel overlaying the canvas) that automatically updates with these scores immediately after an inference pass completes.

Implementation Notes & Resources

• Keep calculations strictly client-side by comparing the flattened 1D arrays of the ground truth and the prediction mask.
• Dice Formula: (2 * overlap) / (total_pixels_mask1 + total_pixels_mask2)
• IoU Formula: overlap / union