Convolutional Neural Network-Based Computer-Assisted Diagnosis of Hashimoto’s Thyroiditis on Ultrasound

Abstract

Purpose This study investigates the efficiency of deep learning models in the automated diagnosis of Hashimoto’s thyroiditis (HT) using real-world ultrasound data from ultrasound examinations by computer-assisted diagnosis (CAD) with artificial intelligence. Methods We retrospectively collected ultrasound images from patients with and without HT from two hospitals in China between September 2008 and February 2018. Images were divided into a training set (80%) and a validation set (20%). We ensembled nine convolutional neural networks (CNNs) as the final model (CAD-HT) for HT classification. The model’s diagnostic performance was validated and compared from two hospital validation sets. We also compared the accuracy of CAD-HT against seniors/junior radiologists. Subgroup analysis of CAD-HT performance in different thyroid hormone levels, such as hyperthyroidism, hypothyroidism, and euthyroidism, was also evaluated. Results 39280 ultrasound images from 21118 patients were included in this study. The accuracy, sensitivity, specificity of the ensemble HT-CAD model were 0.892, 0.890 and 0.895, respectively. HT-CAD performance between two hospitals was not significantly different. The HT-CAD model achieved a higher performance (P < 0.001) when compared to senior radiologists, with a nearly 9% accuracy improvement. HT-CAD had almost similar accuracy (from 0.871 to 0.894) among the three differences of thyroid hormone level subgroups. Conclusion The HT-CAD strategy based on CNN significantly improved the radiologists’ diagnostic accuracy on HT. Our model demonstrates good performance and robustness in different hospitals and for different thyroid hormone levels.

Automated Thyroid Ultrasound Analysis: Hashimoto’s Thyroiditis

“…Convolutional neural networks treat images as input data, analyzing the individual pixels, and aim to achieve a specific classification outcome. 19 Building upon the concept of convolutional neural networks generating feature maps through layered filtering, our C# code implemented a classification evaluation process. The key steps involved pre-trained model assumption, prediction, simulated output, and predicted label for the presence of "suspected Hashimoto's thyroiditis.…”

Investigating the diagnostic efficiency of a computer-aided diagnosis system for thyroid nodules in the context of Hashimoto’s thyroiditis

“…Zhao et al. developed an HT-CAD model based on the convolutional neural network (CNN) with higher diagnostic performance than senior radiologists ( P < 0.001), and the accuracy was improved by nearly 9% ( 30 ). Hou et al.…”

Analyzing the diagnostic value of ultrasound on different-sized thyroid nodules based on ACR TI-RADS

“…On ultrasound, the typical PTMC is solid, hypoechoic/extremely hypoechoic and in vertical position with ill-defined margins and microcalcifications [12][13][14] . As research continues to advance, the diagnostic accuracy of ultrasound has gradually improved, but for the benign nodules in group A, the chronic lymphocytic thyroiditis and granulomatous thyroiditis share obviously overlapped ultrasonic features with PTMC [15][16][17] , and it is also difficult to distinguish follicular adenoma from follicular carcinoma by ultrasound [18] . In the meantime, most nodular goiters are both cystic and solid with clear margins and a transverse diameter that is larger than the longitudinal diameter on ultrasound [19] .…”