The thyroid gland is the largest endocrine organ in the body, and when it behaves abnormally, patient health may suffer. The development of thyroid diagnostic techniques is reviewed considering direct and indirect techniques. Combining spectroscopic approaches with machine deep learning represents a promising diagnostic tool due to its low cost, speed, and good precision. The machine deep learning models for spectral profiling and the evaluation methods are used to test the performance on the bases of precision, sensitivity, specificity, subject operating characteristic (ROC) curve, and F1-score. The analysis of the spectra with machine deep learning models has significant potential for disease diagnosis. [ABSTRACT FROM AUTHOR]