# Documentation for Other MRI Files This document contains information about the OTHER folder, which holds the Jupyter Notebook files I used to create and test models. Each file was used to create more than one model, as simple tweaks to the parameters can lead to more or less efficient models. Model information is located in model_guide.md in the MODELS folder. Knowledge of Python is required to use these files. Pertinent files are listed: * nibabel.ipynb - Testing nibabel. Includes a few utility functions. * nii_to_mosaic.ipynb - Jupyter version of the nii_to_mosaic.py in PROGRAMS. * run_model.ipynb - Jupyter version of the run_model.py in PROGRAMS. * gradcam.ipynb - Jupyter version of the gradcam.py in PROGRAMS. * skullstrip.ipynb - Skullstrips all the .nii files in a folder. * data_validation.ipynb - Lets you view one slice of every scan in a folder, or view every slice of one scan, to validate scan quality. * move*.py - Various small programs for moving files into specific folders. * 3d_image_classification*.ipynb - Template files for training a 3D CNN for classification of lung CT scans. Used as a basis for my own models. * experiment_funcs.py - Contains experimental .nii manipulation functions. Imported by some notebooks, so deleting it will break their imports. #### Models: Alzheimer's * tf-2d-mosaic-dti-AD.ipynb - Models for 2D DTI mosaic. * tf-2d-mosaic-t1-AD.ipynb - Models for 2D T1 mosaic. * tf-3d-cropped-dti-AD.ipynb - Models for cropped 3d DTI scans. * tf-3d-cropped-T1-AD.ipynb - Models for cropped 3D T1 scans. * fastai-2d-mosaic-dti-AD.ipynb - Models for 2D DTI mosaic using the fastai library. Ineffective. Parkinson's * tf-2d-mosaic-Zoomit-PD.ipynb - Models for 2D Zoomit mosaic. * tf-3d-cropped-Zoomit-PD.ipynb - Models for cropped 3D Zoomit scans. * tf_multichannel_2.5d_pd.ipynb - Multi-input models. * tf_multichannel_2d-mosaic-pd.ipynb - Multi-input models designed for use with mosiacs. * tf_transfer* - Transfer learning models. * tf_keras* - Old 3D models with non-normalized T2 scans. Ineffective.