Part 7/11:

• Annotating infected lung regions by radiologists

• Applying active contour algorithms to refine annotations

• Training deep learning models (e.g., U-Net+) on these refined masks

• Generating segmentation maps to highlight inflamed areas

The models did not just detect regions but also quantified the extent of inflammation, which, combined with patient demographics, helped predict whether a patient might need ventilatory support. This integration of quantitative imaging with clinical data exemplifies how AI enhances precision medicine.

Challenges in Medical Image AI

Despite promising results, the session underscored several challenges: