Leveraging Deep Learning to Identify Calcification and Colloid in Thyroid Nodules

Abstract

Background: Both calcification and colloid in thyroid nodules are reflected as echogenic foci in ultrasound images. However, calcification and colloid have significantly different probabilities of malignancy. We explored the performance of a deep learning (DL) model in distinguishing the echogenic foci of thyroid nodules as calcification or colloid.

Methods: We conducted a retrospective study using ultrasound image sets. The DL model was trained and tested on 30,280 images of 1,069 nodules. All nodules were pathologically confirmed. The area under the receiver-operator characteristic curve (AUC) was employed as the primary evaluation index.

Results: The YoloV5 (You Only Look Once Version 5) transfer learning model for thyroid nodules based on DL detection showed that the average sensitivity, specificity, and accuracy of distinguishing echogenic foci in the test group (n=192) was 78.41%, 91.36%, and 77.81%, respectively. The average AUC value of the model was 0.824. The average sensitivity, specificity, and accuracy of the three radiologists were 51.14%, 82.58%, and 61.29%, respectively. The model detection results were better than those of the radiologists.

Conclusions: The study demonstrated that DL performed far better than radiologists in distinguishing echogenic foci of thyroid nodules as calcifications or colloid.

Note:
Funding Declaration: This work was supported by the National Natural Science Foundation of China (No. 82071946), the Zhejiang Provincial Natural Science Foundation of China (No. LZY21F030001 and LSD19H180001) and the Research Program of Zhejiang Provincial Department of Health (No. 2021KY099 and 2022KY110).

Conflicts of Interest: None

Ethical Approval: The research has been carried out in accordance with the World Medical Association Declaration of Helsinki. And this retrospective study was approved by the Ethics Committee of Zhejiang Cancer Hospital, and informed consent was waived (IRB308 2020-287).

Keywords: Thyroid Nodule, ultrasound, Calcification, colloid, Deep Learning