Applying Artificial Neural Networks to Develop a Decision Support Tool for Tis-4N0M0 Non-Small-Cell Lung Cancer...

Area of Science:

• Oncology
• Medical Artificial Intelligence
• Radiation Oncology

Background:

• Evidence comparing surgery and stereotactic body radiation therapy (SBRT) for non-small-cell lung cancer (NSCLC) is limited.
• SBRT offers several potential advantages for NSCLC treatment.

Purpose of the Study:

• To develop and validate artificial neural networks (NNs) for predicting treatment outcomes in NSCLC patients receiving SBRT.
• To aid clinical decision-making by providing accurate outcome predictions.

Main Methods:

• Retrospective analysis of NSCLC patients (Tis-T4N0M0) treated with SBRT (2005-2019).
• Construction and validation of two NNs to predict overall survival (OS) and cancer progression within 5 years post-SBRT.
• Internal and external datasets were used for NN testing, followed by risk group stratification.

Main Results:

• NNs demonstrated strong predictive performance for OS (concordance indexes 0.68-0.76) and cancer progression (AUC 0.70-0.80) across training and test datasets.
• Risk stratification using NNs identified low-risk groups for cancer progression (5.6%-7.0%).
• Approximately 48% of patients with peripheral Tis-4N0M0 NSCLC were identified as low-risk for progression.

Conclusions:

• NN-based outcome predictions are effective for NSCLC patients treated with SBRT.
• These predictive models can guide treatment decisions for both physicians and patients.
• The study opens new possibilities for AI in predicting SBRT outcomes.