Related Experiment Video
Updated: Jun 1, 2025
03:55
Computer-Aided Three-Dimensional Visualization in the Treatment of Locally Advanced Thyroid Cancer
Published on: June 9, 2023
478
Multi-region nomogram for predicting central lymph node metastasis in papillary thyroid carcinoma using multimodal
Shidi Miao1, Qifan Xuan1, Wenjuan Huang2
1School of Computer Science and Technology, Harbin University of Science and Technology, Harbin, China.
Computer Methods and Programs in Biomedicine
|January 19, 2025
Summary
A new multimodal nomogram accurately predicts central lymph node metastasis in papillary thyroid carcinoma (PTC). This model integrates deep learning ultrasound features, CT fat radiomics, and clinical data, outperforming individual assessments.
Area of Science:
- Oncology
- Radiology
- Artificial Intelligence
Background:
- Central lymph node metastasis (CLNM) significantly increases recurrence risk and decreases survival in papillary thyroid carcinoma (PTC).
- Current methods for predicting CLNM in PTC are insufficient, necessitating advanced predictive models.
Purpose of the Study:
- To develop and validate a multimodal and multi-region nomogram (MMRN) for predicting CLNM in PTC.
- To integrate deep learning ultrasound (US) features, computed tomography (CT) fat radiomics, and clinical data for enhanced prediction.
Main Methods:
- A cohort of 661 PTC patients from two centers was analyzed, divided into primary, internal test (ITC), and external test (ETC) cohorts.
- Resnet50 was used for US image analysis, and radiomics features were extracted from CT scans.
- Least absolute shrinkage and selection operator (LASSO) regression was applied for feature selection, with performance validated via five-fold cross-validation.
Main Results:
- The MMRN achieved an Area Under the Curve (AUC) of 0.829 in the ITC and 0.818 in the ETC.
- The model demonstrated good calibration and clinical utility, outperforming radiologist assessments in sensitivity and specificity.
- Inclusion of fat radiomics significantly improved classification accuracy, indicated by Net Reclassification Improvement (NRI) and Integrated Discrimination Improvement (IDI).
Conclusions:
- The developed MMRN is a robust tool for predicting CLNM in PTC, comparable in performance to expert radiologists.
- Fat radiomics features provide valuable supplementary information for CLNM prediction in PTC, enhancing the multimodal model.
Keywords:
Central lymph node metastasisDeep learningMulti-regionMultimodalNomogramPapillary thyroid carcinomaRadiomics
