A Randomized Combined Channel Approach for the Quantification of Color- and Intensity-Based Assays with Smartphones

Quantification of colorimetric assays with smartphones is being increasingly reported. However, a complete characterization of the performance of existing color spaces and single-color channels for optimum color/intensity change quantification is absent. Moreover, it has not been ascertained if it is necessary to utilize existing color spaces to reach optimal assay quantification. In this study, a randomized channel approach was adapted utilizing all single channels from RGB, HSV, and CieLab color space and all nonrepeating random combinations of two and three channels of these color spaces. Assays based on color or intensity change using pH strips and gold or carbon black nanoparticle-containing paper strips were optimized using this approach. Several novel channel combinations showed great promise, in terms of prediction error and interphone variation reduction, outperforming RGB, HSV, and CieLab color spaces. These novel combinations were used in a custom-developed smartphone application that performed automated background subtraction and polynomial regression for the quantification of a lateral flow assay for the detection of goat milk adulteration with cow milk and for pH prediction in soil. For the lateral flow assay the channel combination BSA was found optimum (mean average error = 36% ± 6%; <i>R</i>² = 0.97). For the soil pH assay the channel combination RLC was found optimum (mean average error = 1.31% ± 0.02%; <i>R</i>² = 0.997). The study has shown that nonclassical channel combinations for colorimetric quantification of specific assays are very promising and should be considered for smartphone-based analysis.