In the high-pressure environment of surgical training, effective stress management is crucial for optimal performance and well-being. This study employs mobile EEG technology to monitor cortical indices of stress responses during surgical training tasks. We posit that variations in specific EEG parameters, particularly frequency-specific bandpower and coherence, can offer quantifiable neural markers for stress in surgical settings, as they will reflect self-reported stress levels. Additionally, we have implemented a suite of machine learning classifiers -Support Vector Machines, XGBoost, and Neural Networks- that capitalize on these EEG features. We found a reliable relationship between beta and gamma bandpower and coherence on perceived stress levels. Further, our classifiers demonstrated accuracy rates ranging from 76% to 79% in correctly detecting phases of elevated stress, underscoring the potential of integrating neuroscience and Artificial Intelligence for monitoring mental states in dynamic and demanding contexts.