Advancing thyroid assessment: non-invasive technique and multivariate analysis

Globally, thyroid dysfunction, which includes disorders including hyperthyroidism and hypothyroidism, poses a serious health risk. Patients experience pain and annoyance as a result of the invasive procedures that are frequently used in conventional diagnostic approaches. In this context, the employment of non-invasive techniques like diffuse reflectance spectroscopy (DRS) shows significant promise. A wide spectrum of metabolic problems known as thyroid diseases are defined by deviations in thyroid hormone production and function. Regular assessments are necessary for precise diagnosis and ongoing monitoring of thyroid problems. Many existing diagnostic techniques involve invasive procedures that result in patient discomfort. Identifying non-invasive techniques for predicting thyroid-related parameters is a challenging endeavor. The primary objective of this study is to perform an analysis to evaluate the effectiveness of Standard Normal Variate (SNV) and Partial Least Square Regression (PLSR), in non-invasively predicting thyroid parameters using Diffuse Reflectance Spectroscopy. A cohort of 30 participants was recruited from SRMMCHRC (SRM Medical College Hospital and Research Center) with the approval of the Ethical Committee. These participants were instructed to relax and sit comfortably with informed consent while DRS light in the wavelength range of 550-1040nm was passed through their thyroid region non-invasively. The data was later compared with clinically observed blood tests. In terms of fitting the response variable and estimating Performance metrics, the PLSR (Partial Least Square Regression) model demonstrated an RMSE of 0.286 and 0.93 indicating its strong predictive capability for non-invasive assessment of thyroid parameters. In conclusion, the research showcases the potential of using SNV in combination with the PLSR regression model for the non-invasive assessment of thyroid patients.