Preoperative Diagnosis of Thyroid Nodule Malignancy Using Machine Learning with a Feature Selection Approach

Document Type

Thesis

Degree Name

Master of Science (MS)

Department

Mathematics

Date of Award

Spring 2025

Abstract

Thyroid cancer is prevalent worldwide and presents a significant diagnostic challenge, particularly in distinguishing between malignant and benign nodules within the thyroid gland. Accurate identification of malignant nodules in preoperative stage is crucial for effective clinical treatment. This study used a machine learning approach to develop a predictive model that detects malignant nodules with higher accuracy. We used a comprehensive dataset that includes demographic and clinical information of 1232 nodules collected from 724 patients. Using feature selection techniques, we developed a reduced model after eliminating redundant features. To handle data imbalance, we used Synthetic Minority Oversampling Technique (SMOTE). We also utilized ten-fold cross validation and performance metrics such as accuracy, precision, recall, specificity, and area under the receiver operating characteristic (AUROC) curve to assess the performance of the model. Finally, we conducted a comparative study between our reduced model with the selected features and the full model with all features available in the dataset. The reduced model achieved a more balanced performance compared to the full model.

Advisor

Nahid Hasan

Subject Categories

Mathematics | Physical Sciences and Mathematics

Share

COinS