Results:
This project adapts a regular 3D CNN network seen below:
From the original model the following adaptions were made to take advantage of tensorflow 2.0:
- Using tensorflow probabilities to build a bayesian network, this allows for researchers to get an additional insight into how much the network believes in its final answer.
- Eager executions were used to avoid having to go through the trouble of generating a session runtime for tensorflow
This is not a working product nor does it claim to be. This is merely a proof of concept. The original data was provided by the Parkinson's Progression Markers Initiative.