Smartphone-Based Point-of-Care Testing

Abstract

To enhance sensitivity and accuracy in biosensing applications through the adoption of innovative techniques and technologies, we utilized nitrocellulose paper as the solid phase, providing immobilization force and acting as a filter to reduce background noise. Concurrently, fluorescence particles were used as indicators, and smartphone imaging was employed for data capture. Instead of relying on light intensity measurements, we employed a particle counting method as our optical signals, resulting in improved sensitivity and reliability in various detection applications.Furthermore, we sought to create a portable and cost-efficient platform by utilizing smartphones as the optical signal collectors, eliminating the need for bulky traditional optical equipment. We addressed the limitations of non-competitive immunoassay and immunoagglutination for small-size and one-epitope analytes, like the deadly toxin alpha-amanitin (α-AMA) found in mushrooms. To overcome this challenge, we applied competitive immunoassay in conjunction with our paper-based smartphone detection platform. The particle counting method significantly reduced autofluorescence noises caused by nitrocellulose paper fibers, leading to higher sensitivity. For the quantification of analytes, we encountered difficulties due to sample interferences, such as proteins and debris in salivary samples. To achieve quantitative prediction, we employed three machine learning algorithms (Decision Tree, k-NN, and SVM) to estimate (-)-trans-delta-Tetrahydrocannabinol (THC) concentration in human saliva by training the machine learning models with various diluted saliva samples. In addition to quantification, classification is another significant function of machine learning. We utilized the eXtreme Gradient Boosting (XGBoost) algorithm to identify the dominance bacterial species in mixture samples. Creative conjugation of quorum sensing-related peptides with fluorescence particles improved sensitivity in traditional immunoaggregation assays. The non-specific peptides-set effectively identified bacterial species. To further enhance our biosensors, we utilized the autofluorescence properties of cells for free-label detection. We analyzed soil microbiome components using smartphone images under various color filters. This approach, in combination with XGBoost, enabled efficient and specific identification of bacterial species. By integrating novel technologies and machine learning algorithms into our biosensing platforms, we achieved significant advancements in sensitivity, accuracy, and versatility. These developments have the potential to revolutionize diagnostic and biosensing applications, paving the way for rapid and reliable point-of-care testing in various fields.