Feb. 2026
Synopsis¶
A machine-learning coursework project focused on classifying mammographic breast density into the four BI-RADS density categories using a subset of the RSNA mammography dataset. The project combined wavelet-based image preprocessing with transfer learning, CNN architecture comparison, subgroup performance analysis, and inspection of the model’s learned feature embeddings.
Skills¶
Below are the skills and experiences gained from the project:
Experience with subgroup analysis and model evaluation using ROC curves and AUC
Experience with model and embedding inspection using PCA and t-SNE
Competency with PyTorch and PyTorch Lightning
Experience applying transfer learning and comparing CNN architectures
Competency with PyWavelets for wavelet-based image denoising
Competency with OpenCV for filtering, image fusion, and contrast enhancement
