Improved evaluation of oral premalignant lesions (OPLs) at risk of malignant transformation is critical to improving oral cancer outcomes. Towards this goal, a multimodal optical imaging system (MMIS) that integrates white-light, autofluorescence (AF), and high-resolution microendoscopy (HRME) imaging to non-invasively evaluate OPLs was developed. The MMIS uses an automated, real-time image analysis algorithm to classify imaged sites as low- or high-risk for high-grade dysplasia or cancer. To analyze HRME images containing regions without quality information, a novel algorithm utilizing deep learning and probabilistic methods was also developed. Finally, two clinical studies were conducted to evaluate the MMIS. In the first study, high-risk OPL patients under surveillance for malignant progression received biopsies if clinically indicated, or if indicated by the MMIS evaluation, at the clinician’s discretion. The results demonstrated that the addition of the MMIS improved identification of high-grade dysplasia, compared to clinical evaluation alone. The second study explored the potential of the MMIS in a community dental setting, where most lesions are benign confounders. The study found that AF+HRME had a higher diagnostic accuracy than the clinical impression of a group of dental residents, and that the addition of AF and HRME to the residents’ clinical impressions improved their diagnostic performance. Together, these studies indicate the potential of the MMIS to improve the clinical evaluation of OPLs in multiple settings.