Explainable Medical AI / Computer Vision

X-GastroAI: explainable gastric screening

A clinician-facing AI pipeline for gastric cancer screening that pairs prediction with interpretable visual explanations.

Why explainability mattered

Medical AI systems cannot stop at a class label. For screening workflows, the model also needs to show why it produced a prediction, where the visual signal came from, and how a user should interpret confidence.

X-GastroAI was built around that principle: combine deep learning inference with Grad-CAM saliency maps so predictions are paired with a visual explanation layer.

Technical highlights

• ResNet-50 classifier for gastric cancer screening support.
• Grad-CAM overlays to make model attention visible and interpretable.
• Streamlit inference UI for a fast clinician-facing prototype experience.
• Real-time prediction flow focused on usability, confidence, and model transparency.

What I took away

The project reinforced a simple idea: model accuracy is only part of the product. In high-trust workflows, the surrounding system - input handling, explanation, UI clarity, and deployment ergonomics - determines whether the AI can actually be used.

• Medical AI
• Computer Vision
• ResNet-50
• Grad-CAM
• Streamlit
• Explainable AI