/Parkinsons-DL-Classifier

A classifier created to detect PD on patients using fMRI data

Primary LanguageJupyter NotebookMIT LicenseMIT

Parkinsons-DL-Classifier

Architecture-of-the-Inception-V3-module-41

About

A Deep Learning approach utilizing a CNN based on Inception V3 model, for classifing fMRI scans to Parkinson's Disease group or Healthy group. This model was developed as part of a semester project in the Mobile & Electronic Health Technologies course @ NTUA 2021-2022.

Data

For this project the database ds000245 from OpenNeuro was used.

The data were converted from the initial .nii format to .png format using the converter that can be found here. After conerting the images to png format some data cleaning was performed and the resulting data split is the following:

Healthy PD
Train 170 170
Validation 34 34
Test 17 17

Score - Confusion Matrix

Accuracy Precision Recall F1
Score 64.7% 58.6% 100% 36.9%
Healthy PD
Healthy 17 12
PD 0 5