This project contains detailed analysis and comparison of 3 different methods for solving problem of fundus retinal vessel segmentation. It was made as final project for the 6-th semester Computer Science in Medicine subject course at Poznan University of Technology.
All code was merged as part of report in eye_fundus.ipynb
. Inside 3 methods of approaching this problem are outlined:
rf
- Ridge Filtering - Using CV image-related functions and ridge filtersml
- Machine Learning - Using well-established machine learning classifiers for image chunksdl
- Deep Learning - Using Deep Neural Network with U-net architecture
References:
- Liskowski, Krawiec 2016 - Segmenting Retinal Blood Vessels With Deep Neural Networks
- HRF Dataset - https://www5.cs.fau.de/research/data/fundus-images/