The goal of this project is to build a model to identify signs of Alzheimer's disease in brain scans of patients,
developing also a way of explain to doctors which parts of the scan are most relevant to the decision.
I structured it as a 3-part tutorial for those who would like to replicate it:
- Using two balance approach: Sick vs nonSick and Data augmentation (SMOOTE)
- Testing various CNN models (ex. ResNet,VGG16, MobileNet...)
- GradCAM interpretation to show doctors which parts of the brain are most interesting in determining the presence of the disease