Use of Fuzzy Logic in the Classification of Thyroid Nodules Detected by Ultrasonography

Abstract

Abstract Objective Employ fuzzy logic to auxiliary in diagnosis and malignancy grade of thyroid nodules by ultrasound. Methods A cross-sectional study evaluating 75 exams results from patients with a thyroid nodule. The following ultrasound findings were evaluated employing a quantitative score: not suspicious, not very suspicious, moderately suspicious, and highly suspicious. The echographic features evaluated for suspicion of malignancy were based on the following nodule components: composition, echogenicity, shape, margin, and echogenic foci, graded using the Thyroid Imaging Data and Reporting System by the American College of Radiology. By combining ultrasound scoring and the Bethesda System for Reporting Thyroid Cytopathology using fuzzy logic, a classification for thyroid nodules was constructed. Results Hypoechogenicity and microcalcifications were the findings that showed the best interaction with malignancy on ultrasound, while shape and margin showed the smallest estimation errors when compared with composition. A classification for thyroid nodules was suggested based on the 95% confidence interval of hypoechogenicity and microcalcifications: not suspicious (< 24.6); not very suspicious (24.6–48.0); moderate (48.1–64.5); moderately suspicious (64.6–77.0); highly suspicious (77.1–92.7); and malignant (> 92.7). Conclusion By fuzzy logic, a classification for thyroid nodules diagnosed by ultrasound supported by echogenicity and nodular microcalcifications was constructed with a simple practical application.