Deep-learning project for renal pathology prognosis, connected to Gabriele’s master’s thesis and later Django-based clinical decision-support work.

The pipeline combines whole-slide renal biopsy images with immunofluorescence images to estimate five-year kidney failure risk. It tests WSI-only, fluorescence-only, and fused WSI plus fluorescence settings, then evaluates whether combining modalities improves the deep learning prognostic score.

The repository documents training scripts, feature extraction, linear-classifier experiments, split analysis, and evaluation metrics such as accuracy, AUC, F1-score, precision, recall, and specificity. The project is useful as an applied example of medical-image preprocessing, multimodal learning, and model evaluation under limited clinical data.

This work is related to the later ESKD Prediction Web Application, which extends the clinical prediction direction into a deployed Django application for AI-assisted risk estimation.

---

← Back to projects