"Gray zone" thyroid follicular tumors are difficult to diagnose, especially when distinguishing between benign follicular thyroid adenoma (FTA) and malignant carcinoma (FTC). Thus, proper classification of thyroid follicular diseases may improve clinical prognosis. In this study, the diagnostic performance of metabolite enzymes was evaluated using imaging mass spectrometry to distinguish FTA from FTC and determine the association between metabolite enzyme expression with thyroid follicular borderline tumor diagnosis. Air flow-assisted desorption electrospray ionization mass spectrometry imaging (AFAIDESI-MSI) was used to build a classification model for thyroid follicular tumor characteristics among 24 samples. We analyzed metabolic enzyme marker expression in an independent validation set of 133 cases and further evaluated the potential biological behavior of 19 thyroid borderline lesions. Phospholipids and fatty acids (FAs) were more abundant in FTA than FTC (p < 0.001). The metabolic enzyme panel, which included FA synthase and Ca2+-independent PLA2, was further validated in follicular thyroid tumors. The marker combination showed optimal performance in the validation group (area under the ROC, sensitivity, and specificity: 73.6%, 82.1%, and 60.6%, respectively). The findings indicate that AFAIDESI-MSI, in combination with low metabolic enzyme expression, could play a role in the diagnosis of thyroid follicular borderline tumors for strict follow-up.