Machine Learning Enables Reliable Colorimetric Detection of pH and Glucose in Wearable Sweat Sensors

Abstract In healthcare, blood pH and glucose levels are critical indicators, especially for chronic conditions like diabetes. Although taking blood samples is accurate, it is invasive and unaffordable for many. Wearable sensors offer non‐invasive and continuous detection methods, yet face major challenges, such as high cost, inaccuracies, and complex interpretation. Colorimetric wearable sensors integrated with machine learning (ML) are introduced for accurately detecting pH values and glucose concentrations in sweat. These battery‐free and cost‐effective biosensors, made of cotton textiles, are designed to work seamlessly with smartphones for data collection and automated analysis. A new pH indicator is synthesized with enhanced sensitivity and two types of glucose sensors are developed by depositing enzymatic solutions onto cotton substrates. The sensors' performance is assessed using standard solutions with known pH levels ranging from 4 to 10 and glucose concentrations between 0.03 to 1 m m . The photos captured from these sensors are then analyzed by image processing and three different ML algorithms, achieving an accuracy of 90% in pH and glucose detection. These findings provide effective synthesis methods for textile‐based sweat sensors and demonstrate the significance of employing different ML algorithms for their colorimetric analysis, thus eliminating the need for human intervention in the process.