Improving The Diagnosis of Thyroid Cancer by Machine Learning and Clinical Data

Nan Miles Xi,Lin Wang,Chuanjia Yang

DOI: https://doi.org/10.48550/arXiv.2203.15804

2022-03-28

Abstract:Thyroid cancer is a common endocrine carcinoma that occurs in the thyroid gland. Much effort has been invested in improving its diagnosis, and thyroidectomy remains the primary treatment method. A successful operation without unnecessary side injuries relies on an accurate preoperative diagnosis. Current human assessment of thyroid nodule malignancy is prone to errors and may not guarantee an accurate preoperative diagnosis. This study proposed a machine framework to predict thyroid nodule malignancy based on a novel clinical dataset we collected. The 10-fold cross-validation, bootstrap analysis, and permutation predictor importance were applied to estimate and interpret the model performance under uncertainty. The comparison between model prediction and expert assessment shows the advantage of our framework over human judgment in predicting thyroid nodule malignancy. Our method is accurate, interpretable, and thus useable as additional evidence in the preoperative diagnosis for thyroid cancer.

Machine Learning,Quantitative Methods,Applications

What problem does this paper attempt to address?

The paper aims to improve the diagnostic accuracy of thyroid cancer through machine learning and clinical data. Specifically, the researchers collected a new clinical dataset containing 1,232 nodules from 724 patients and trained six advanced machine learning models to predict the malignancy of thyroid nodules. By evaluating model performance using 10-fold cross-validation, Bootstrap analysis, and permutation predictor importance methods, the study results show that these machine learning models outperform human experts in predicting thyroid nodule malignancy, particularly in terms of accuracy and sensitivity. Additionally, the study identified several variables crucial for predicting the malignancy of thyroid nodules, such as calcification, blood flow richness, and nodule location. These findings help improve the accuracy and reliability of preoperative diagnosis of thyroid cancer, thereby reducing unnecessary surgical risks and the possibility of postoperative recurrence.