Machine learning models for diagnosing lymph node recurrence in postoperative PTC patients: a radiomic analysis.

Area of Science:

• Oncology
• Radiology
• Medical Imaging

Background:

• Postoperative papillary thyroid cancer (PTC) patients frequently present with enlarged cervical lymph nodes, complicating recurrence assessment.
• Inflammation or hyperplasia can cause lymph node enlargement, mimicking metastasis.

Purpose of the Study:

• To evaluate the diagnostic performance of computed tomography (CT) imaging and radiomic analysis in identifying cervical lymph node recurrence in patients with PTC after surgery.
• To develop a predictive model for cervical lymph node recurrence risk.

Main Methods:

• Retrospective analysis of 194 PTC patients post-thyroidectomy.
• Delineation of Regions of Interest (ROIs) on CT images for 693 lymph nodes (302 positive, 391 negative).
• Extraction of radiomic features using Python and development of Lasso, SVM, and RF radiomic models; creation of a nomogram combining clinical data and radiomic scores.

Main Results:

• Identified 35 significant radiomic features from CT imaging.
• Developed and validated Lasso, Support Vector Machine (SVM), and Random Forest (RF) radiomic models for recurrence detection.
• Assessed model efficacy using ROC curves, calibration curves, and Decision Curve Analysis (DCA).

Conclusions:

• Radiomic analysis of CT imaging shows promise in differentiating recurrent cervical lymph nodes in postoperative PTC patients.
• The developed nomogram integrating clinical factors and radiomic features can aid in predicting recurrence risk.
• These findings support the potential of radiomics as a non-invasive tool for postoperative PTC surveillance.