Attention Deficit Hyperactivity Disorder (ADHD) is a neurobiological condition with genetic origins. Identifying this disorder is a complex process, as other neurobiological disorders can manifest symptoms similar to those observed in patients with ADHD. Faced with this diagnostic challenge, the search for innovative approaches has intensified in recent years. Advances in the field of neuroimaging, including functional Magnetic Resonance Imaging (fMRI), have played a fundamental role in improving clinical decision-making. fMRI is a technique that allows the measurement of brain activity through a sequence of three-dimensional images collected at specific time intervals, creating a 4D file that includes spatial dimensions (x, y, z) and the temporal dimension (t). In this context, this project proposes the use of a specialized framework for processing and manipulating these images in order to prepare them for the architecture required by the 3D Convolutional Neural Network (3D CNN). This approach aims to identify Attention Deficit Hyperactivity Disorder, seeking to contribute to the advancement of the current diagnostic process. The choice of the 3D CNN model is based on its ability to explore the three-dimensional characteristics of fMRI data, allowing for a deeper and more accurate analysis of the complex brain connections associated with the images.
The evaluation metrics used in the model achieved as baseline 77% "Binary Accuracy", 100% "Precision", and a "Recall" of 25%.