| Summary: | Integrating artificial intelligence with SARS-CoV-2 diagnostics can help in the timely execution of pandemic control and monitoring plans. To improve the efficiency of the diagnostic process, this study aims to classify fluorescent images via traditional machine learning and deep learning-based transfer learning. A previous study reported a CRISPR-Cas13a system combined with total internal reflection fluorescence microscopy (TIRFM) to detect the existence and concentrations of SARS-CoV-2 by fluorescent images. However, the lack of professional software and excessive manual labor hinder the practicability of the system. Here, we construct a fluorescent image dataset and develop an AI-boosted CRISPR-Cas13a and total internal reflection fluorescence microscopy system for the rapid diagnosis of SARS-CoV-2. Our study proposes Fluorescent Images Classification Transfer learning based on DenseNet-121 (FICTransDense), an approach that uses TIRF images (before and after sample introduction, respectively) for preprocessing, including outlier exclusion and setting and division preprocessing (i.e., SDP). Classification results indicate that the FICTransDense and Decision Tree algorithms outperform other approaches on the SDP dataset. Most of the algorithms benefit from the proposed SDP technique in terms of Accuracy, Recall, F1 Score, and Precision. The use of AI-boosted CRISPR-Cas13a and TIRFM systems facilitates rapid monitoring and diagnosis of SARS-CoV-2.
|
|---|