Machine learning-based segmentation of lateral brain ventricles in MRIs using Keras
This contains one example pipeline of generating training "patches" of a 3D MRI, labelling those patches as part of the lateral ventricle or not, desining a CNN model in Keras, training it on aforementioned patches, testing the model on seprate MRIs, and evaluating the accuracy of the model's prediction of where in the MRI the ventricles are. Other CNNs were desgined, trained, and tested but their architectures are not shown here. Simple ITK, scikit-learn, Keras were all used.