This study, by Amar et al,. compared the effectiveness of a baseline approach to the use of a Multi-Headed Deep Learning (DL) CNN-based approach to diagnose ADHD (Attention Deficit Hyperactivity Disorder) using EEG (electoencephalography) data, widely used in ADHD research due to its accessibility and convenience. The Baseline Approach involved training one deep learning model using EEG signals from two channels without any enhancements. The Multi-headed approach, however, used several DL models that worked together to process input data for a proper diagnosis. The dataset collected data from 79 participants, which consisted of 42 healthy adults and 37 adults diagnosed with ADHD. The signals were recorded across four states, including resting, not resting, listening to sounds, and during cognitive challenges. Additionally, the dataset consisted of four.mat files containing FC (control group women), MC (control group men), FADHD (ADHD group women), MADHD (ADHD group men), and each file was arranged as a 1 x 11 cell array, with each cell providing details such as the number of subjects, signal samples, and EEG channels. A neural network was also proposed to transform the output vector into a probability distribution across classes. To evaluate the results, five performance metrics were used to assess the models: accuracy (ACC), precision (Prec), recall (Rec), F1 score, and the area under the ROC (Receiver Operating Characteristic) curve (AUC). To efficiently process sequential data, like EEG signals, the baseline and Multi-headed approach both used recurrent neural network architectures, like Bidirectional Long Short-Term Memory (BI-LSTM), Long Short-Term Memory (LSTM), and Gated Recurrent Unit (GRU). Across configurations, the Multi-headed approach performed better compared to the Baseline Approach, especially due to LSTM and BI-LSTM models, with an accuracy of 89.87%, despite both models having a lower accuracy alone. Moreover, the Multi-headed approach outperformed the Baseline Approach across the other metrics: Prec, ACC, Rec, F1, and AUC; this improved identification reflected fewer false positives/negatives and a higher discriminatory power for the Multi-headed method. The study explored the classification of healthy participants, and those with ADHD.