Toward Clinical Translation of a Bimodal Fluorescence Tool for Oral Cancer Detection and Biopsy Guidance

Early- and large-scale screening and diagnosis of a disease such as oral cancer is a boon for the underdeveloped population. Our present study focuses on clinical translation of a bimodal fluorescence imaging and spectroscopy-based portable device with the potential for early and real-time diagnosis of oral cancer. This smartphone-based device collects fluorescence images and spectra from oral tissue sites within a few minutes and can assist clinicians in biopsy guidance with its advanced contrast-enhancing algorithm between normal and abnormal tissue regions. A total of 184 patients and volunteers were tested and classified using an integrated multilayer perceptron (MLP) classification scheme, in which discriminative features were extracted from both spectral and imaging data. The combined multimodal algorithm achieved strong performance, with a macro-average precision of 92.67%, sensitivity of 91.88%, F1-score of 91.93%, specificity of 95.89%, and overall accuracy of 94.59%.