Unveiling the Anomalies: Using Autoencoders for Detecting Structural Changes in Huntington's Disease from MRI Images

Huntington's disease (HD) is a severe neurodegenerative condition characterized by the gradual worsening of motor skills, cognitive abilities, and mental health. Detecting and tracking the disease early is crucial for prompt treatment and enhancing patient prognosis.Traditional diagnostic methods, such as clinical assessments and genetic testing, often fail to detect subtle structural brain abnormalities until the disease is in an advanced stage.This study investigates the use of autoencoders, a type of artificial neural network, to identify and quantify structural anomalies in MRI images of individuals with HD. By training autoencoders on MRI data from HD patients and healthy controls, we aim to enhance the sensitivity and specificity of early detection methods for HD-related structural changes. MRI brain scans were collected, resized to 256x256 pixels for computational efficiency, and normalized. The dataset was divided into training, validation, and test sets, following a 70/15/15 ratio. The autoencoder model used consists of an encoder that reduces the input images into a latent space representation, while the decoder rebuilds the images from this compressed form. The model achieved an accuracy of 95%, demonstrating its ability to effectively differentiate between MRI images of HD patients and controls based on structural anomalies.