At its core, Chiron utilizes a Convolutional Neural Network (CNN) machine learning model to assess the user’s likelihood of having Parkinson’s disease based on their spiral and wave drawings with 87% accuracy. These drawings are a unique and non-invasive way to capture subtle motor impairments associated with Parkinson’s disease. However, before being trained and tested, a series of important steps are taken. First, the dataset is carefully refined and filtered to ensure its quality. Then, the data is augmented, a technique of artificially increasing the dataset by creating modified copies of existing data. Finally, the CNN model is trained and then tested on this dataset of spirals and waves to ensure a high level of diagnostic accuracy. This code is run on a virtual machine in Oracle Cloud.

A Convolutional Neural Network is a type of deep learning algorithm well-suited for image recognition. It is made up of multiple layers, including convolutional layers, pooling layers, and fully connected layers. In convolutional layers, filters are applied to the input image to extract features such as edges, textures, and shapes which are then passed through pooling layers, which are used to down-sample the feature maps, reducing the spatial dimensions while retaining the most important information. The output of the pooling layers is then passed through the fully connected layers, which are used to make a prediction or classify the image. In this case, it returns a value of either 0 for healthy images or 1 for images showing signs of Parkinson’s.

While the backend CNN model was coded using Python, the frontend design of Chiron was coded using Swift. The frontend code sends a request to an API containing an image of the spiral or wave and is returned a result of either being healthy or showing signs of Parkinson’s.