Automated Lymph Node and Extranodal Extension Assessment Improves Risk Stratification in Oropharyngeal Carcinoma.

Area of Science:

• Oncology
• Medical Imaging
• Artificial Intelligence

Background:

• Extranodal extension (ENE) is a key prognostic biomarker in oropharyngeal carcinoma (OPC).
• Diagnosis of ENE traditionally requires surgical pathology, limiting its use in pretreatment settings.
• Accurate pretreatment risk stratification is crucial for optimizing OPC management.

Purpose of the Study:

• To evaluate the prognostic value of AI-predicted ENE node number in OPC.
• To assess the ability of an AI imaging platform to integrate lymph node autosegmentation with ENE prediction.
• To determine if AI-ENE improves risk stratification compared to existing models.

Main Methods:

• A retrospective study of 1,733 OPC patients treated with definitive radiation therapy across three institutions.
• Utilized a deep learning model for lymph node autosegmentation and a subsequent model for ENE prediction on pretreatment CT scans.
• Analyzed associations between AI-predicted ENE node number (AI-ENE) and distant control (DC) and overall survival (OS) using multivariable Cox regression.
• Assessed improvements in risk stratification by incorporating AI-ENE into RTOG-0129 and AJCC-8th edition staging.

Main Results:

• AI-ENE node number was independently associated with poorer DC (HR 1.44) and OS (HR 1.30) (P < .001).
• Increasing AI-ENE node number showed incremental association with worse outcomes, particularly DC.
• Incorporating AI-ENE improved C-indices for both OS and DC when added to RTOG-0129 groupings and AJCC-8 staging (P < .001).
• Risk-stratification improvements were most significant in HPV-negative patients.

Conclusions:

• Automated AI-ENE node number is a novel and significant risk factor in OPC.
• AI-ENE offers a valuable tool for enhancing pretreatment risk stratification and clinical decision-making in OPC.
• This AI approach may help personalize treatment strategies for oropharyngeal cancer patients.