Detection of cancer cells over blood microscopy images based on shape anomaly

The codes in this repository have been implemented in MATLAB.

• The aim of this project is detecting white blood cancer cells based on the collected images by pattern recognition and machine learning techniques.

• The project has 6 main topics:

1. Image acquisition
2. Image preprocessing
3. Image Segmentation
4. Morphologic Operations
5. Feature Extraction
6. Clasification

• In this project the different results have been shown with different methods and their impact on classification accuracy.
  • CIELAB Lab color segmentation and K-Means clustering used for image segmentation,
  • Watershed algorithm applied to detect and separate overlapped white blood cells.
  • Texture, statistical and geometrical features are used for feature extraction of white blood cells
  • Support Vector Machine and Multilayer Perceptron Neural Network used for classification.

In total 108 images were analyzed and up to 95% accuracy has been achieved.

Dataset can be request in below link

https://homes.di.unimi.it/scotti/all/