In vivo identification of the parathyroid glands during surgery using diffuse reflectance spectroscopy

Background Accurate identification of the parathyroid glands during neck endocrine surgery is essential for optimal patient outcomes; however this remains challenging, even for experienced surgeons. Diffuse reflectance spectroscopy (DRS) is an optical technique able to provide information on tissue morphology, function and biochemical composition. Previous ex vivo study has demonstrated that DRS can differentiate parathyroid from thyroid tissue with an accuracy of 93.3%. This study aims to assess if DRS can accurately identify parathyroid glands in the in vivo setting. Method This is a prospective in vivo feasibility study; patients undergoing neck endocrine surgery were prospectively recruited. Specimens were analysed in vivo with a sterilised handheld DRS probe and spectral data collected. Binary classification was performed with the XGB supervised machine learning classifier; this was evaluated with stratified 5-fold cross-validation. Results Of 26 patients recruited, 22(85%) were female, with a median age of 49(18–79) years. A total of 1472 spectra were collected across 21 normal parathyroids, 15 parathyroid adenomas, 25 fat samples and 8 thyroid samples. In differentiating normal parathyroids from thyroid and fat, the classifier achieved a diagnostic accuracy of 98.26%. In differentiating parathyroid adenomas from thyroid and fat, the accuracy was 97.52%. The system distinguished normal parathyroid tissue from adenomas with an accuracy of 99.58%. Conclusion This study demonstrates that DRS can intra-operatively differentiate parathyroid glands from other tissues in the neck, and can distinguish adenomatous glands. Further study is justified in the form of clinical trials to assess the impact of utilising this technology on patient outcomes.