Artificial intelligence-based algorithm for predicting outcomes in early-stage lung cancer: An annotation-free...

Area of Science:

• Oncology
• Radiology
• Artificial Intelligence

Background:

• Surgery is the standard treatment for stage I non-small cell lung cancer (NSCLC).
• Conventional prognostic factors for NSCLC are subjective and variable.
• Objective prediction methods are needed for stage I NSCLC prognosis.

Purpose of the Study:

• To develop and validate an annotation-free artificial intelligence (AI) model for prognosis prediction in stage I NSCLC.
• To integrate computed tomography (CT) imaging and clinical data for enhanced predictive accuracy.
• To improve risk stratification and support personalized treatment planning.

Main Methods:

• Developed an AI algorithm to predict pathological classifications from CT and clinical data.
• Refined and validated the model using multi-institutional prospective trial data and a validation cohort.
• Trained models to predict 5-year disease-free and overall survival, evaluating performance using AUC.

Main Results:

• Models integrating CT imaging and clinical data outperformed those using single data types.
• The AI model achieved an AUC of 0.787 for pathology prediction.
• Combined AI, clinical, and CT assessment data yielded the highest AUC for 5-year survival prediction (0.757 for DFS, 0.756 for OS).

Conclusions:

• Annotation-free AI models integrating CT and clinical data offer accurate, objective prognosis prediction for stage I NSCLC.
• These AI models complement conventional diagnostics.
• AI-driven predictions support personalized multidisciplinary treatment planning for NSCLC.