Prome, Tasnim AhsanTasnim, AnikaShahat, Md. Shajjat HossainTamim, Fardin HassanSagor, Al-AminMayabi, Simran ZamanBiswas, Smita2026-08-102026-08-1020262026-01ID 24141244ID 23341111ID 24141225ID 24141242ID 24241240https://hdl.handle.net/10361/28885This thesis is submitted in partial fulfillment of the requirements for the degree of Bachelor of Science in Computer Science, 2026.Cataloged from PDF version of thesis.Includes bibliographical references (pages 40-41).The vast advancement of neurodiagnostic technology offers an opportunity to transform mental health assessment with more precise, objective evaluations. Which previously was reliant on subjective matters like individual perspectives (Doctors in this case), interpretations, and such. With the advancement of technology in the field of medical science, there are so many resources that can be used and worked on to find efficient and better solutions for any disease or health issue. Assessing the state of one’s mental health will have an impact on unleashing one’s potential fully. This research solely focuses on the application of deep learning-driven MRI processing in neurodiagnostic frameworks to improve mental health assessment. In this research, our main focus was on Attention Deficit Hyperactivity Disorder (ADHD), which is a very common and concerning mental health issue, especially amongst young children. To understand its mechanism, it’s crucial to investigate how functional connectivity patterns between various brain regions portray the neurobiological processes of the condition. With the help of the Graph Attention Network with Phenotype fusion (GAT-PhoNet) method, a deep learning model, we analyze the brain activity pattern through graphs, which is created from the ADHD-200 dataset resting-state fMRI data and phenotypic data that include age, gender, and IQ information. The model aimed to identify the brain activity patterns at the network level while also obtaining the crucial clinical covariates. Our GAT-PhoNet model achieved a classification accuracy of 72.09% along with an AUC-ROC score of 0.7517 and F1-scores of 0.74 for healthy controls and for ADHD subjects 0.68. This proved to be an understandable and powerful technique, which helps with ADHD detection through objective methods. The study demonstrates how multimodal graph-based deep learning systems can enhance neurodiagnostic procedures by providing more accurate quantitative analysis results.51 pagesen-USAttribution-NonCommercial-NoDerivatives 4.0 InternationalBRAC University theses are protected by copyright. They may be viewed from this source for any purpose, but reproduction or distribution in any format is prohibited without written permission.http://creativecommons.org/licenses/by-nc-nd/4.0/Neurodiagnostic technologyMental health assessmentDeep learningGraph neural networksAttention-deficit hyperactivity disorderADHDfMRI dataMagnetic resonance imagingDiseases detectionNeuroimagingImage processingMedical imagesMagnetic resonance imaging--Data processing.Image analysis.Brain--Magnetic resonance imaging.Deep learning (Machine learning).Attention-deficit hyperactivity disorder--Diagnosis.Diagnostic imaging--Data processing.Neural networks (Computer science).Thesis