Related Experiment Video
Updated: Oct 14, 2025
03:55
Computer-Aided Three-Dimensional Visualization in the Treatment of Locally Advanced Thyroid Cancer
Published on: June 9, 2023
680
Computed Tomography-Based Radiomics Model to Predict Central Cervical Lymph Node Metastases in Papillary Thyroid
Jingjing Li1,2, Xinxin Wu2, Ning Mao3,4,5
1Second Clinical Medicine College, Binzhou Medical University, Yantai, China.
Frontiers in Endocrinology
|November 8, 2021
Summary
This study developed a computed tomography (CT)-based radiomics model to predict central lymph node metastases (CLNM) in papillary thyroid carcinoma (PTC) patients. The combined model integrating radiomic and clinical factors demonstrated effective preoperative prediction of CLNM.
Area of Science:
- Radiology
- Oncology
- Medical Imaging
Background:
- Papillary thyroid carcinoma (PTC) is the most common type of thyroid cancer.
- Central lymph node metastases (CLNM) are frequent in PTC and impact treatment decisions.
- Accurate preoperative prediction of CLNM is crucial for optimal surgical planning.
Purpose of the Study:
- To develop and validate a computed tomography (CT)-based radiomics model for predicting central lymph node metastases (CLNM) in patients with papillary thyroid carcinoma (PTC).
- To assess the performance of the radiomics model, both alone and combined with clinical risk factors, for preoperative CLNM prediction.
- To evaluate the clinical usefulness of the developed model.
Main Methods:
- Retrospective analysis of 678 PTC patients with preoperative CT scans (plain and contrast-enhanced).
- Extraction of radiomics features, followed by feature selection using ANOVA and LASSO algorithm.
- Development of radiomics models using various machine learning algorithms (e.g., linear-SVM) and a combined model incorporating clinical risk factors (sex, age, tumor diameter, CT-reported lymph node status).
- Performance evaluation using ROC curves, calibration curves, and decision curve analysis (DCA).
Main Results:
- 14 radiomic features and 4 clinical variables were significantly associated with CLNM.
- The combined radiomics model using linear-SVM achieved AUCs of 0.747 (training), 0.710 (internal test), and 0.764 (external test).
- The combined model demonstrated superior sensitivity and accuracy compared to an experienced radiologist in the internal test set.
- Favorable agreement between predicted and actual CLNM probabilities was observed, and DCA confirmed clinical usefulness.
Conclusions:
- A CT-based radiomics model, incorporating radiomic features and clinical risk factors, can effectively predict CLNM in PTC patients preoperatively.
- This non-invasive model offers a valuable tool for improving preoperative assessment and surgical planning for PTC.
- The combined radiomics approach shows promise for enhancing diagnostic accuracy beyond traditional methods.
