Abstract Background: Accidental excision or damage to a healthy parathyroid gland (PG) during thyroidectomy and the inability to identify diseased PGs during parathyroidectomy can result in postsurgical complications. Meanwhile, hyperparathyroidism includes primary and secondary hyperparathyroidism, which can cause damage in skeleton, kidney, and heart, with bone pain and renal colic. Therefore, it is critical to distinguish healthy parathyroid when the operation is being processed since the extent of the parathyroidectomy depends on the accuracy of the diagnosis. Methods: Human cancerous and healthy thyroid tissues were obtained from Shanghai General Hospital. Porcine thyroid glands, recurrent laryngeal nerves (RLN), adipose tissues, muscle tissues, lymph nodes, and parathyroid glands were harvested in fresh. Each sample was divided into two sections, one for CARS imaging and the other for hematoxylin and eosin staining (H&E). The H&E images were then classified by a senior pathologist. Some imaging analysis methods were performed to quantify some parameters, such as overall normalized intensity, regional normalized intensity, etc., to quantitatively differentiate the six types of tissues. Deep-learning algorithms were developed for auto-diagnosing thyroid cancer and for differentiating parathyroids, thyroids, lymph nodes, fats, RLN, and muscles. Results: Human thyroid cancer tissue can be identified immediately and automatically through CARS imaging combined with deep-learning algorithms. Parathyroid adenoma/hyperplasia presented as enlarged chief cells and clear cells with reduced fat droplets compared with normal parathyroid. The CARS identification results were confirmed by an experienced pathologist. All six different porcine tissues can be identified by the same system, and the correspondence to human equivalents was confirmed by the pathologist. The accuracy of both identification is above 90%. Quantitative analysis was performed to differentiate parathyroid, thyroid, lymph node, and fat, and differentiate RLN from muscle. Significance: The intelligent-augmented label free imaging system combining CARS imaging and deep-learning algorithm can reliably identify thyroid cancer tissue and differentiate it from thyroid gland, lymph node, and adipose tissue morphologically, without staining, accurately, and immediately. Therefore, this system is a potential tool for quantitatively differentiating parathyroid glands and RLN from surrounding tissues and identifying thyroid cancer tissue in real-time for diagnosis and during parathyroid surgery. Ultimately, the imaging technique can be incorporated into handheld probe for use in clinical diagnosis, such as cancer margin detection during surgery for minimal dissection and reducing nerve impairment during prostate cancer surgery. Citation Format: Jiasong Li, Ye Wang, Yunjie He, Mary C. Farach-Carson, Stephen T.C. Wong. 3D label-free, real-time, intelligent-augmented chemistry-sensitive imaging to identify parathyroid adenoma and hyperplasia and to classify parathyroid glands and recurrent laryngeal nerve during surgery [abstract]. In: Proceedings of the Annual Meeting of the American Association for Cancer Research 2020; 2020 Apr 27-28 and Jun 22-24. Philadelphia (PA): AACR; Cancer Res 2020;80(16 Suppl):Abstract nr 4252.