A Comparison of Different Radiomics Methods Predicting Cervical Lymph Node Metastasis in Papillary Thyroid Carcinoma

Area of Science:

• Radiology
• Oncology
• Medical Imaging

Background:

• Accurate preoperative identification of cervical lymph node metastasis in papillary thyroid carcinoma (PTC) is crucial for surgical planning.
• Ultrasound-based radiomics offers potential for non-invasive prediction of metastasis.

Purpose of the Study:

• To develop and compare the performance of handcrafted radiomics, deep learning radiomics, and a combined model for predicting cervical lymph node metastasis in PTC.
• To evaluate the efficacy of these models in guiding surgical treatment decisions.

Main Methods:

• Retrospective analysis of 441 PTC patients.
• Extraction of handcrafted radiomics features (physician-selected) and deep learning radiomics features (automated DenseNet121).
• Development of a combined model integrating both feature types; performance assessed using ROC analysis and DeLong's tests.

Main Results:

• The combined model demonstrated superior AUC in the training set (0.790) compared to handcrafted (0.743) and deep learning (0.730) models.
• While the combined model showed higher AUC in the testing set (0.761), statistical significance was not reached.
• Handcrafted radiomics achieved high accuracy (0.714 training, 0.707 testing), outperforming deep learning radiomics (0.698 training, 0.662 testing).

Conclusions:

• A combined radiomics model utilizing conventional ultrasound images improves the prediction of cervical lymph node metastasis in PTC.
• This integrated approach offers enhanced predictive performance over individual radiomics models for clinical application.