DentiVueX: A Low-Cost Intraoral Spectral Imaging Platform for Oral Health Assessment

Abstract

Oral diseases affect approximately 3.5 billion people worldwide, yet the diagnostic tools capable of providing meaningful spectral tissue and bacterial assessment remain too expensive and physically incompatible with the intraoral environment for widespread clinical use. This thesis presents DentiVueX — a low-cost, portable intraoral imaging platform that extends a smartphone-based multispectral reconstruction framework previously developed at the University of Washington's Biophotonics and Imaging Laboratory to a miniature intraoral camera probe. The system supports two imaging modalities: white-light hyperspectral reflectance imaging and UV-excited autofluorescence imaging for bacterial porphyrin detection. Linear and nonlinear Wiener Matrix estimation pipelines are implemented and evaluated across multiple optical hardware configurations and calibration chart sizes. The nonlinear model, incorporating gamma decoding, polynomial feature expansion, brightness-weighted least squares, and Tikhonov regularization, consistently outperforms the linear baseline across all tested conditions. The optimal hyperspectral configuration achieves a nonlinear reconstruction RMSE of approximately 0.050 and GFC of 0.972, with skin-tone-adjacent patches reaching a mean GFC of 0.993. The optimal autofluorescence configuration achieves RMSE of approximately 0.004 and GFC of 0.989. These results establish the feasibility of RGB-based spectral reconstruction from a low-cost intraoral camera and provide a foundation for future clinical validation and translation.