Hyperspectral imaging and artificial intelligence to detect oral malignancy – part 1 - automated tissue classification of oral muscle, fat and mucosa using a light-weight 6-layer deep neural network

Abstract

Background Hyperspectral imaging (HSI) is a promising non-contact approach to tissue diagnostics, generating large amounts of raw data for whose processing computer vision (i.e. deep learning) is particularly suitable. Aim of this proof of principle study was the classification of hyperspectral (HS)-reflectance values into the human-oral tissue types fat, muscle and mucosa using deep learning methods. Furthermore, the tissue-specific hyperspectral signatures collected will serve as a representative reference for the future assessment of oral pathological changes in the sense of a HS-library. Methods A total of about 316 samples of healthy human-oral fat, muscle and oral mucosa was collected from 174 different patients and imaged using a HS-camera, covering the wavelength range from 500 nm to 1000 nm. HS-raw data were further labelled and processed for tissue classification using a light-weight 6-layer deep neural network (DNN). Results The reflectance values differed significantly (p < .001) for fat, muscle and oral mucosa at almost all wavelengths, with the signature of muscle differing the most. The deep neural network distinguished tissue types with an accuracy of > 80% each. Conclusion Oral fat, muscle and mucosa can be classified sufficiently and automatically by their specific HS-signature using a deep learning approach. Early detection of premalignant-mucosal-lesions using hyperspectral imaging and deep learning is so far represented rarely in in medical and computer vision research domain but has a high potential and is part of subsequent studies.

Comparison of Hyperspectral Imaging and Microvascular Doppler for Perfusion Monitoring of Free Flaps in an In Vivo Rodent Model

“…In this context, the use of artificial intelligence may help to simplify the interpretation of the values and increase their significance. In a previous study, a deep learning algorithm supported HSI was successfully tested [ 58 ], which could be transferred to the MFF monitoring in the near future.…”

Hyperspectral imaging for early diagnosis of diseases: A review

“…ROI marking can be done manually by pathologists and physicians themselves for accurate results. Weijtmans et al, 2019, Blanco et al, 2012., and Thiem et al, 2021…”

“…Blanco et al, 2012 propose using the NIR hyperspectral imaging technique for the classification of kidney stones in a faster and more robust way for the urologist for rapid diagnosis using deep learning. The use of an H2O flow model was proposed in Thiem et al, 2021's work for detecting oral malignancy by performing tissue classification. While Zarei et al, 2017 presented a methodology for automated detection of prostate glandular structures and their nuclei from a Tissue Micro Array M. H. F. Aref et al, 2021 gave a proposal for a non‐invasive custom optical imaging system for detection of arm blood vessels, both studies using K‐means for clustering and segmentation.…”

“…ROI marking can be done manually by pathologists and physicians themselves for accurate results. Weijtmans et al, 2019, Blanco et al, 2012., and Thiem et al, 2021 use ROIs for better classification and detection of tumours. In the study by Abdlaty et al, 2021, ROIs are marked manually before processing images.…”

“…Further improvements can include using a more elaborate network like CNN. Thiem et al, 2021 proposed in their paper classification of HS reflectance values into human oral tissue types: fat, muscle, and mucosa. At least 5 ROIs were placed on the specimen, and samples underwent histological evaluation for pre‐processing data.…”