attention deficit hyperactivity disorder (ADHD) is a prevalent public health issue that impacts individuals globally. Characterized by symptoms such as inattentiveness, hyperactivity, and impulsivity, ADHD often persists throughout life, significantly affecting an individual's social, educational, and occupational functioning. It is frequently associated with various mental health challenges, including disruptive behaviors, emotional dysregulation, and an increased risk of self-harm, emphasizing the importance of early and accurate diagnosis. This paper presents a diagnostic approach leveraging Functional Magnetic Resonance Imaging (fMRI) enhanced with optical amplification for ADHD detection. By utilizing Convolutional Neural Networks (CNNs), this method extracts essential features from fMRI data to improve diagnostic accuracy. The study further explores the efficacy of three optimization algorithms—Maximum Adaptive Moment Estimation (AdaMax), Accelerated Nesterov Adaptive Moment Estimation (Nadam), and Root Mean Square Propagation (RMSProp)—to refine classification outcomes. Experimental results demonstrate that RMSProp yields the highest accuracy at 98.33%, surpassing leading architectures such as ResNet (95.83%) and GoogleNet (93.55%). These findings mark a significant advancement in automated ADHD diagnosis, offering a robust, high-accuracy method that could streamline clinical assessments and provide earlier intervention opportunities to mitigate long-term adverse effects associated with the disorder.