Thorax-encompassing multi-modality PET/CT deep learning model for resected lung cancer prognostication: A...

Area of Science:

• Oncology
• Medical Imaging
• Artificial Intelligence

Background:

• Early-stage non-small cell lung cancer (NSCLC) patients often undergo surgery, but recurrence remains a significant risk.
• Conventional cancer staging, while useful, may not fully capture prognostic information present in medical images.
• There's a need for advanced models integrating diverse data to improve risk stratification for post-surgical NSCLC patients.

Purpose of the Study:

• To develop and evaluate a deep learning model (DLM) for predicting NSCLC recurrence-free survival (RFS).
• To assess if integrating FDG PET/CT imaging with clinical, surgical, and pathological data improves RFS prediction and risk stratification compared to conventional staging.

Main Methods:

• Retrospective analysis of surgically resected NSCLC patients from two institutions (2009-2018).
• A multi-modal DLM was developed using preoperative FDG PET/CT imaging and clinical, surgical, and pathological data.
• The DLM's performance in predicting RFS and stratifying risk was evaluated on testing and external validation cohorts, compared against conventional staging.

Main Results:

• The multi-modal DLM demonstrated significant predictive value for RFS in both testing (AUC=0.78) and external validation (AUC=0.66) cohorts.
• The DLM effectively stratified patients into high, medium, and low-risk groups (p < 0.001), outperforming conventional staging which failed to stratify patients.
• The DLM also provided significant sub-risk stratification within conventional stages I and II.

Conclusions:

• This study introduces the first multi-modal DLM using imaging and clinical data to predict RFS in post-surgical NSCLC patients.
• The DLM significantly improves risk stratification compared to conventional staging and can identify patients at higher risk within established stages.
• This model holds potential to aid clinicians in identifying NSCLC patients who may benefit from additional adjuvant therapy.