Hello Juuso,
I understand that you want to know whether it would be beneficial to try traditional machine learning methods like Partial Least Squares Regression (PLS) + Support Vector Machines (SVM) before trying CNNs for the classification of 3D MRI data.
In the context of MRI scans, the suitability of PLS for classification can be limited as PLS might struggle to capture the informative textures and complex patterns which are crucial in medical imaging. Although PLS reduces the dimensionality of the data, it may not capture the spatial and volumetric information effectively unlike CNNs that are specifically designed to capture spatial hierarchies and patterns in image data.
However, Partial Least Squares Regression can be adapted for classification tasks, including image classification, by using it as a feature extraction or dimensionality reduction method before applying a classifier like SVM on the extracted features. The function plsregress in MATLAB is designed for regression problems with continuous outcomes. For categorical outcomes, PLS can be adapted for classification by using the scores from PLS as input features for a classifier such as SVM with the 'fitcecoc' function.
Please refer to the below documentation to learn more about:
- 'fitcecoc' function: https://www.mathworks.com/help/stats/fitcecoc.html
- 'plsregress' function: https://www.mathworks.com/help/stats/plsregress.html
Hope it helps!
