Major depressive disorder (MDD) is a widespread mental illness affecting millions worldwide. Timely and accurate diagnosis is crucial for effective treatment, yet conventional methods often depend on subjective evaluation. This study presents a novel method for detecting MDD through the analysis of electroencephalogram (EEG) signals. We utilize Maximum Likelihood Estimation (MLE) for selecting EEG channels, aiming to enhance the performance of machine learning classifiers. Comprising EEG recordings from known MDD patients and healthy individuals, the dataset enables an objective comparison. The effectiveness of the proposed method was tested using a Multi-Layer Perceptron Neural Network (MLPNN) classifier. Our results show that incorporating MLE for channel selection significantly improves classification accuracy when compared to using MLPNN alone. This highlights MLE as a valuable approach to refine the detection of MDD through machine learning models. Our findings demonstrate that EEG signals can successfully distinguish between individuals with MDD and those without. Additionally, the use of MLE significantly boosts the classifier’s performance. The improved results obtained by combining MLE and MLPNN suggest the potential for broader applications in MDD detection. Future work will focus on exploring additional classification algorithms, investigating other feature selection techniques, and employing alternative channel selection methods to further enhance model accuracy. This research provides valuable insights into the use of EEG signals and machine learning for more reliable mental health diagnostics.