Computerized cytometry and wavelet analysis of follicular lesions for detecting malignancy: A pilot study in thyroid cytology

详细信息    查看全文

• 作者：Hayim Gilshtein ; MD^a ; ^{hgilshtein@gmail.com} ; Michal Mekel ; MD^a ; ^c ; Leonid Malkin ; MD^b ; Ofer Ben-Izhak ; MD^b ; ^c ; Edmond Sabo ; MD^b ; ^c
• 刊名：Surgery
• 出版年：2017
• 出版时间：January 2017
• 年：2017
• 卷：161
• 期：1
• 页码：212-219
• 全文大小：252 K
• 卷排序：161

文摘

The cytologic diagnosis of indeterminate lesions of the thyroid involves much uncertainty, and the final diagnosis often requires operative resection. Computerized cytomorphometry and wavelets analysis were examined to evaluate their ability to better discriminate between benign and malignant lesions based on cytology slides.MethodsCytologic reports from patients who underwent thyroid operation in a single, tertiary referral center were retrieved. Patients with Bethesda III and IV lesions were divided according to their final histopathology. Cytomorphometry and wavelet analysis were performed on the digitized images of the cytology slides.ResultsCytology slides of 40 patients were analyzed. Seven patients had a histologic diagnosis of follicular malignancy, 13 had follicular adenomas, and 20 had a benign goiter. Computerized cytomorphometry with a combination of descriptors of nuclear size, shape, and texture was able to predict quantitatively adenoma versus malignancy within the indeterminate group with 95% accuracy. An automated wavelets analysis with a neural network algorithm reached an accuracy of 96% in identifying correctly malignant vs. benign lesions based on cytology.ConclusionComputerized analysis of cytology slides seems to be more accurate in defining indeterminate thyroid lesions compared with conventional cytologic analysis, which is based on visual characteristics on cytology as well as the expertise of the cytologist. This pilot study needs to be validated with a greater number of samples. Providing a successful validation, we believe that such methods carry promise for better patient treatment.