A code developed in Python is used for the classification of MRI images from a dataset sourced from OASIS obtained from Kaggle (https://www.kaggle.com/datasets/ninadaithal/imagesoasis). For classification, a CNN is constructed using keras, which has also been used in Kaggle code for this dataset (see link).
- First, the images are processed to have a single black-and-white channel.
Two approaches are followed:
- In the first approximation, for both binary and multiclass classification, the number of images from the class with the lowest number of images will be selected.
- In the second approximation, for multiclass classification, oversampling will be performed using SMOTE.