Cognitive impairments are commonly considered as comorbidities in various mental disorders (MDs), such as Attention Deficit-Hyperactivity Disorder and Bipolar Disorder. Assistive technology and Serious Games (SGs), in particular, have been widely used to monitor and treat cognitive deficits, since they are considered effective. In our previous study, we introduced the design and development of an educational, brain-based SG, namely MindOfMine, which presented promising results in the cognitive enhancement of people who are diagnosed with MDs. This study explores the integration of multimodal data such as Electroencephalogram, Skin Conductance, Body Temperatarure, Game Logs and Keystrokes in order to develop a closed-loop SG platform (SGP), by integrating open access datasets. Our analysis present promising results in both disorder detection and prediction of required number of Sessions (accuracy: 0.875 in 3-class Disorder detection and 0.893 in a 6-session prediction using RUBoost). By developing a closed loop SGP that combine the positive nature of SGs with AI, we foresee SGs applications that automate the diagnosis of MDs, predict required therapy sessions, and personalizes treatment plans, thereby enhancing therapeutic outcomes and streamlining clinical workflows.