Application of Deep Learning for Breast Cancer Metastasis Detection in Sentinel Lymph Node Biopsies

Author: Manuel Cossio

Director: Dr. Laura Igual

Link to dataset

It's necessary to download the dataset and make it available in the Google Drive directory where the Colab notebooks are going to be executed.

Dataset SNL

Sections

1. Data exploratory analysis

Data exploratory analysis notebook

This notebook provides the code for the data exploratory analysis.

2. Classification with hand-crafted features and shallow classifiers

Hand-crafted features classification notebook

This notebook provides the code to:

• Produce the handcrafted features
  • Color histograms
  • Mean brightness
• Classify with linear classifiers
  • Support Vector Classifier
  • Stochastic Gradiend Descent
• Classify with non-linear classifiers
  • Random Forest Regressor

3. Classification with Deep Learning

Deep Leaning classification notebook

This notebook provides the code to:

• Data augmentation
• Outlier extraction
• Classification with transfer learning
• Explainability analysis
  • Automatic feature extraction
  • Gradient-Weighted Class Activation Mapping (Grad-CAM)

Experiments

Data and plots of the experiments conducted can be found in the following link.

Experiments_plots

Consulted resources

Complete beginner's guide

CNN Nasnet Mobile

Baseline Keras CNN - ROC - FAST

ML pipeline