Artificial Intelligence based facial recognition for Mood Charting among men on life style modification and it's correlation with cortisol.

Today, clinicians and researchers believe that mood disorders in children and adolescents remain one of the most under diagnosed mental health problems. Mood disorders in adolescents also put them at risk for other conditions that may persist long after the initial episodes of depression are resolved. In our study we have assessed the mood state spectrum of a person over the time and validated the same by correlating with salivary cortisol, psychologist assessment results.
Methods and Materials: Images from the training dataset are classified according to one among the nine emotions. The Images that are classified accordingly are taken as training set and a suitable convolutional neural network is trained/retrained for this data. The mean values predicted moods are considered as input for another model that predicts higher the stress level.
Results: With the inception v3 trained for 1,00,000 times with the data set that's close to 12,000 images classified accordingly to the nine emotion classes as specified by psychologist the model was able to obtain 78.4% of testing accuracy while with the near perfect training accuracy.
Conclusion: The mood analysis was conclusively helpful in the estimation of the negative emotion parameter as close to the values that are obtained by with Depression, Anxiety and stress scale (DASS21) the mood analysis. The salivary cortisol as unbiased variable correlating with DASS 21 score could have a potential uses in early detection of mood disorder and correction. This platform will be helping the subject as a early mood screening tool.
(Copyright © 2019 Elsevier B.V. All rights reserved.)