Identifying Oral Cancer Using Multispectral Snapshot Camera

Multispectral imaging has emerged as an effective tool for detecting malignant changes affiliated with cancers. The added spectral dimension enables the visualization of underlying biochemical information in tissues. Unfortunately, existing acquisition systems which use multispectral filter wheels or variable light source for spectral data collection, face the challenge of co-registration and are hard to be realized in real-time as acquisition time per frame is high. A Multispectral SnapShot Camera (MSSC) is demonstrated to address these obstacles, which is capable of capturing data simultaneously in multiple wavelengths with a single click. This brings down the total acquisition time to a few milliseconds per data cube (spatial+spectral) which encourages translation of this technology for real-time diagnosis. The MSSC is composed of 16 spectral bands with peak wavelengths in the visible range 460 nm - 600 nm. The application of MSSC in the field of oral cancer diagnosis has been demonstrated with in-vivo tissue images captured from 10 patients with oral dysplasia. Principal Component Analysis (PCA) and Spectral Angle Mapper (SAM) was used to differentiate between healthy and malignant tissue. PCA was used to compress the multispectral datacube to extract wavelengths with maximum variance and SAM discriminates between tissue conditions based on distinct spectral profiles. The mean absorption spectrum plotted for them shows a trend of higher absorption peak value in malignant tissue regions than in healthy regions, thereby demonstrating the capability of MSSC for tumor tissue classification.