Automated Calculation of Liquid Crystal Sensing Images Based on Deep Learning

Liquid crystal (LC)-based sensors have been extensively applied in the detection of chemical and biological events. However, the calculation of the optical images of the LC-based sensors is usually time-consuming and also might bring some errors due to the use of different judgment criteria by different users. In the present study, an automated calculation method for LC sensing images based on deep learning is provided. A convolutional network is trained with the prepared LC sensing images and their corresponding segmentation annotations to predict the positive responses. The ratio is calculated from the area of positive response to the total area selected by our image processing method. The robustness of the proposed algorithm is validated on both the test set and the label-free Cd²⁺ detection. The results show that the method based on deep learning can detect the positive response area in real time and the speed is much faster than the manual processing method. In addition, deep learning method can be directly applied to other label-free molecular detection assays.