Classifying SPECT scans of Parkinson's Disease which are available from the PPMI repository. You can open the notebook file in Goolge Colab directly from here:
Note that due to the Colab markup, the notebook does not properly display in GitHub.
Classifying SPECT scans of Parkinson's Disease which are available from the PPMI repository.
This work pertains to unpublished research and is not publicly available before acceptance with a journal. Therefore, the notebook cannot be run by unauthorized users, and does not reveal any new material.
You can open the notebook file in Goolge Colab directly from here:
Licensed under this license.
This notebook shows how we performed the experiment to fine-tune the Inception V3 classifier to distinguish patients with and without signs of Parkinson's disease.
The notebook is optimized to work with Google Colab. It hides all code by default so that you can run it sequentially from top to bottom. If you want to see the implementation, just double-click into any cell, and it will open in a side-by-side view.
This code is part of the publication Automatic classification of dopamine transporter SPECT: deep convolutional neural networks can be trained to be robust with respect to variable image characteristics.
The data are a derivative of the DAT scans available from the PPMI repository. They were processed to represent the central 5 slices of the putamen in one slice by averaging them. For details, please refer to the paper.
The data was then split randomly into a training and a validation set. As we tested the performance on an independant test set drawn from clinical routine which cannot be published, this notebook does not contain testing of the trained classifier.
The data as used in the publication can be downloaded here:
If you want to run the notebook from Google Colab, put the data into your Google Drive, and adapt the path below in the respective cell.
It is adviced to copy the Colab notebook to your own Google Drive before connecting to a runtime an running the code, though it is not strictly required. Note, however, that if you don't copy it, any changes you make, and also all results, will be discarded when you close the notebook.
The data (under the ./data/ path in the repository) will be downloaded and unzipped automatically. In case you are running the notebook in a hosted runtime, the download will be temporary.