Regarding features for model building

[Arshdeep Kaur]

There are a lot of variables in TADPOLE_D1_D2.csv i want to train a model using LSTM with considering each patients data at 5 visits(bl,m06,m12,m18,m24) . By training the model using each patients data at 4 previous visits and predicting the DX label at current visit. There are two types of freesurfer pipeline generated data(longitudnal  UCSFFSL and cross-sectionl UCSFFSX

 .......so can i use patients demographic , and cross-sectional freesurfer MRI data for prediction. I am very new to this field. Please help

[Neil Oxtoby]

I suggest that you explore feature selection methods.

Regarding FreeSurfer, you can use either XS or longitudinal values. It's up to you. The FreeSurfer wiki discusses the difference: surfer.nmr.mgh.harvard.edu/fswiki/LongitudinalProcessing.