Related Experiment Video
Updated: Jun 17, 2026
11:17
Thoracoscopic Extended Right Middle Plus Lower Sleeve Lobectomy for Non-Small-Cell Lung Cancer
Published on: February 27, 2026
AI-Driven Preoperative Chest Radiograph Analysis for Prognostic Stratification in Surgically Resected Pathological
Hyun Joo Shin1, Eun Hye Lee2, Se Hyun Kwak2
1Department of Radiology, Research Institute of Radiological Science and Center for Clinical Imaging Data Science, Yongin Severance Hospital, Yonsei University College of Medicine, Seoul, South Korea.
Journal of Imaging Informatics in Medicine
|June 16, 2026
Summary
Artificial intelligence (AI) analysis of preoperative chest radiographs (CXRs) can predict outcomes for early-stage non-small cell lung cancer (NSCLC) patients. This AI-driven approach shows promise in enhancing prognostic accuracy for surgical candidates.
Area of Science:
- Oncology
- Radiology
- Artificial Intelligence
Background:
- Early-stage non-small cell lung cancer (NSCLC) requires accurate prognostic prediction for optimal surgical management.
- Current prognostic models often rely on clinical variables, pathologic tumor size, and CT parameters.
- The potential of artificial intelligence (AI) in analyzing preoperative chest radiographs (CXRs) for outcome prediction remains an area of active investigation.
Purpose of the Study:
- To investigate the utility of AI-driven analysis of preoperative CXRs in predicting postoperative outcomes for patients with early-stage NSCLC.
- To compare the performance of AI-based prognostic models with traditional models using clinical and imaging variables.
Main Methods:
- Retrospective analysis of 416 patients with pathological stage 1 NSCLC who underwent curative surgical resection.
- AI analysis of preoperative CXRs was performed to detect abnormalities, generating an abnormality score.
- Cox proportional hazards regression was used to identify predictors of recurrence-free survival (RFS), comparing AI-based models with clinical and CT-based models.
Main Results:
- AI detectability and AI abnormality score were significant independent risk factors for poor RFS.
- An AI-based prognostic model demonstrated comparable performance (c-index 0.795) to models based on pathologic tumor size or CT parameters (c-index 0.794).
- Incorporating the AI abnormality score into CT-derived models significantly improved discrimination (c-index 0.837 vs. 0.821).
Conclusions:
- AI-driven analysis of preoperative CXRs can effectively predict postoperative prognosis in early-stage NSCLC patients.
- AI analysis of CXRs offers a valuable tool to enhance preoperative risk stratification and surgical decision-making.
- This AI approach shows potential to improve prognostic accuracy beyond traditional methods.
