"Validation of Our Approach."

Our PoC has been validated and has achieved exceptional accuracy outperforming previous tests. One of our key analyses was conducted on an open-source fNIRS dataset (the ‘fNIRS2MW dataset’ from Tufts University) generated by an experiment in which 68 subjects were exposed to multiple tasks with different stress levels. We applied Neural Networks and Data Science techniques, such as label mapping, dynamic normalization, feature selection, and sigmoid, achieving a peak classification accuracy of 98.7% finding strong correlations between stress, physical and cognitive performance.    

These results have the seed of an app incorporating a base AI model, data-collecting interface and a performance dashboard that identifies brain or heart activity associated with a range of physical and mental outcomes, including injury, recovery from mental or physical trauma and general ‘wellness’.

Our software has a broad range of applications, most notably in supporting healthcare professionals preventing serious health conditions in their patients like ischemic strokes, where emotional stress is a significant trigger element. We also applied our models to sports datasets, revealing a correlation between stress and injuries in football players.  

PoC code and software entails a set of Jupyter notebooks that perform signal analysis, feature selection and apply several ML models and corresponding cross-validation for hyperparameter optimization, including neural networks.