Colorimetric assays are an important tool in point-of-care testing that offers several advantages to traditional testing methods such as rapid response times and inexpensive costs. A factor that currently limits the portability and accessibility of these assays are methods that can objectively determine the results of these assays. Current solutions consist of creating a test reader that standardizes the conditions the strip is under before being measured in some way. However, this increases the cost and decreases the portability of these assays. The focus of this study is to create a machine learning algorithm that can objectively determine results of colorimetric assays under varying conditions. To ensure the flexibility of a model to several types of colorimetric assays, three models were trained on the same convolutional neural network with different datasets. The images these models are trained on consist of positive and negative images of ETG, fentanyl, and HPV Antibodies test strips taken under different lighting and background conditions. A fourth model is trained on an image set composed of all three strip types. The results from these models show it is able to predict positive and negative results to a high level of accuracy.
Experimental Design: In part one of this study, exercise-induced eccrine sweat was collected from 50 healthy individuals and analyzed using mass spectrometry, protein microarrays, and quantitative ELISAs to identify a broad range of proteins, antibody isotypes, and cytokines in sweat. In part two of this study, cortisol and melatonin levels were analyzed in exercise-induced sweat and plasma samples collected from 22 individuals.
Results: 220 unique proteins were identified by shotgun analysis in pooled sweat samples. Detectable antibody isotypes include IgA (100% positive; median 1230 ± 28 700 pg/mL), IgD (18%; 22.0 ± 119 pg/mL), IgG1 (96%;1640 ± 6750 pg/mL), IgG2 (37%; 292 ± 6810 pg/mL), IgG3 (71%;74.0 ± 119 pg/mL), IgG4 (69%; 43.0 ± 42.0 pg/mL), and IgM (41%;69.0 ± 1630 pg/mL). Of 42 cytokines, three were readily detected in all sweat samples (p<0.01). The median concentration for interleukin-1α was 352 ± 521 pg/mL, epidermal growth factor was 86.5 ± 147 pg/mL, and angiogenin was 38.3 ± 96.3 pg/mL. Multiple other cytokines were detected at lower levels. The median and standard deviation of cortisol was determined to be 4.17 ± 11.1 ng/mL in sweat and 76.4 ± 28.8 ng/mL in plasma. The correlation between sweat and plasma cortisol levels had an R-squared value of 0.0802 (excluding the 2 highest sweat cortisol levels). The median and standard deviation of melatonin was determined to be 73.1 ± 198 pg/mL in sweat and 194 ± 93.4 pg/mL in plasma. Similar to cortisol, the correlation between sweat and plasma melatonin had an R-squared value of 0.117.
Conclusion: These studies suggest that sweat holds more proteomic and hormonal biomarkers than previously thought and may eventually serve as a noninvasive biomarker resource. These studies also highlight many of the challenges associated with monitoring sweat content including differences between collection devices and hydration, evaporation losses, and sweat rate.