Artificial Intelligence Role in Subclassifying Cytology of Thyroid Follicular Neoplasm

• DOI: 10.31557/apjcp.2023.24.4.1379
• Publication Date: 2023-04-28T20:03:37Z
• Authors: Mona Alabrak; Mohammad Megahed; Asmaa Alkhouly; Ammar Mohammed; Habiba Elfandy; Neveen Tahoun; Hoda Ismail
• Keywords: Thyroid neoplasm; Neoplasm; Thyroid Nodules

Objective: Fine needle aspiration cytology has higher sensitivity and predictive value for diagnosis of thyroid nodules than any other single diagnostic methods. In the Bethesda system reporting thyroid, category IV, encompasses both adenoma carcinoma, but it is not possible to differentiate lesions in practice can be only differentiated after resection. this work, we aim at exploring ability a convolutional neural network (CNN) model sub-classifying cytological images IV into follicular carcinoma. Methods: We used cohort cases n= 43 with extracted 886 train CNN aiming sub-classify neoplasm (Bethesda IV) either or Result: our study, subclassification (n = 28/43, n 527/886) from carcinoma 15/43, 359/886), achieved an accuracy 78%, 88.4%, specificity 64% area under curve (AUC) score 0.87 each Conclusion: Our high recognizing amongest smears follciualr neoplasms, thus as ancillary technique subcalssification Iv smears.