CORTISOL HORMONE AND STRESS LEVELS: COLORIMETRIC ASSESSMENT WITH ARTIFICIAL INTELLIGENCE SUPPORTED IMAGE PROCESSING METHOD

Cingöz Çapan E., Çapan M. E., Öncel H. U., Öncel H. U., Arıcan E.

Konya Mühendislik Bilimleri Dergisi, cilt.14, ss.70-96, 2026 (TRDizin)

Özet

This study presents a new method for determining cortisol hormone levels, a key biomarker of stress, using microfluidic pads to collect sweat samples. The pads facilitate the colorimetric detection of cortisol levels via the blue tetrazolium method. The resulting color change is analytically assessed using Convolutional Neural Networks (CNN), Decision Trees, and Vector Regression, alongside advanced image processing techniques. The developed algorithm is robust, providing reliable results despite hardware variations and color distortion, enhancing the system's applicability and generalizability across different environments. Validation studies conducted with ELISA and a colorimeter revealed that the system achieved an accuracy of 84.2% in determining users' cortisol levels. Additionally, psychosocial stress levels were assessed using the Copenhagen Psychosocial Risk Assessment and the Perceived Stress Scale tests during the collection of sweat samples from 20 participants. The results demonstrated a significant correlation between cortisol levels and stress, confirming the method's reliability and effectiveness in various applications.