Node-RADS in cervical cancer: a multi-reader agreement and diagnostic performance study

Area of Science:

• Radiology
• Oncology
• Medical Imaging

Background:

• Accurate preoperative lymph node staging is crucial for cervical cancer treatment planning.
• The Node Reporting and Data System (Node-RADS) 1.0 aims to standardize lymph node assessment on MRI.

Purpose of the Study:

• To evaluate the inter-reader agreement and diagnostic performance of Node-RADS for cervical cancer lymph node staging.
• To assess the impact of reader experience on Node-RADS performance.

Main Methods:

• Retrospective analysis of 439 cervical cancer patients with preoperative MRI and lymph node dissection.
• Four readers (2 senior, 2 junior) independently applied Node-RADS scores, blinded to pathology.
• Agreement assessed using kappa statistics; diagnostic performance evaluated against histopathology.

Main Results:

• Senior readers demonstrated near-perfect agreement (kappa > 0.95) for Node-RADS scores and nodal status.
• Junior readers showed moderate to substantial agreement (kappa 0.554-0.754).
• Senior readers had higher diagnostic performance (AUC 0.868) for nodal metastasis detection compared to junior readers (AUC 0.758).

Conclusions:

• Node-RADS provides reliable lymph node staging in cervical cancer with good inter-reader agreement, especially among experienced radiologists.
• Reader experience influences diagnostic performance, highlighting the need for training and standardization.