Deep learning-enabled fluorescence imaging for surgical guidance: in silico training for oral cancer depth quantification

Oral cancer surgery requires accurate margin delineation to balance complete resection with post-operative functionality. Current fluorescence imaging systems provide two-dimensional margin assessment yet fail to quantify tumor depth prior to resection. Harnessing structured light in combination with deep learning (DL) may provide near real-time three-dimensional margin detection. A DL-enabled fluorescence spatial frequency domain imaging (SFDI) system trained with tumor models was developed to quantify the depth of oral tumors. A convolutional neural network was designed to produce tumor depth and concentration maps from SFDI images. Three representations of oral cancer lesions were developed to train the DL architecture: cylinders, spherical harmonics, and composite spherical harmonics (CSHs). Each model was validated with SFDI images of patient-derived tongue tumors, and the CSH model was further validated with optical phantoms. The performance of the CSH model was superior when presented with patient-derived tumors ( ). The CSH model could predict depth and concentration within 0.4 mm and , respectively, for tumors with depths less than 10 mm. A DL-enabled SFDI system trained with CSH demonstrates promise in defining the deep margins of oral tumors.