Development of a prediction model for recurrent parathyroid carcinoma lesions based on 3D-EnCT and ultrasound imaging features

Abstract

Purpose: This study aimed to analyze the three-dimensional enhanced computed tomography (3D-EnCT) and ultrasound imaging features of recurrent parathyroid carcinoma lesions and develop a prediction model based on these features.

Methods: The clinical data of 34 patients (48 cases) with recurrent parathyroid carcinoma who underwent surgical treatment at Beijing Chaoyang Hospital's Thyroid and Neck Surgery Department between January 2017 and April 2024 were retrospectively analyzed. A total of 103 suspicious lesions were identified through a combination of preoperative 3D-EnCT and ultrasound examinations. Patients admitted prior to 1 January 2023 were included in the training set, and those admitted after 1 January 2023 were included in the validation set. In the training set, lesions were categorized as positive or negative based on pathological analysis. Statistically significant imaging features were identified via intergroup comparisons. An imaging prediction model was developed based on the 3D-EnCT and ultrasound features, and the predictive performance of the model was evaluated via receiver operating characteristic curves in the validation set.

Results: Arterial- and venous-phase CT values, lesion boundaries, and blood flow signals were associated with pathological positivity. The 3D-EnCT prediction model based on these features achieved areas under the curve (AUCs) of 0.9 and 0.714 in the training and validation sets, respectively, whereas the ultrasound prediction model achieved AUCs of 0.601 and 0.621, respectively. The 3D-EnCT model demonstrated superior predictive performance.

Conclusion: The 3D-EnCT prediction model demonstrated superior predictive performance for recurrent parathyroid carcinoma lesions.

Keywords: 3D-EnCT; Prediction model; Recurrent parathyroid carcinoma; Ultrasound.