2024 World Conference on Lung Cancer (WCLC) - Posters-P3.08E.02 An Artificial Intelligence Driven Approach to Predict Pathological Response to Neoadjuvant Chemoimmunotherapy for Non-Small Cell Lung Cancer

P3.08E.02 An Artificial Intelligence Driven Approach to Predict Pathological Response to Neoadjuvant Chemoimmunotherapy for Non-Small Cell Lung Cancer

Back to course

Pdf Summary

This study investigates the use of artificial intelligence (AI) to predict the pathological response to neoadjuvant chemoimmunotherapy in patients with Non-small Cell Lung Cancer (NSCLC). Currently, evaluating pathological responses before surgery is challenging. A retrospective study involved 314 patients with stages II to III NSCLC who received combined anti-programmed death-1 (PD-1) therapy and chemotherapy across five medical centers in China. The study was conducted from May 2019 to August 2023.<br /><br />Patients were split into three groups: a training set (n=202), an internal validation set (n=51), and an external validation set (n=61). Radiomic features were extracted from chest CT scans before and after therapy to calculate 'delta features,' which captured changes in imaging characteristics. The Mann-Whitney U test was used to filter irrelevant features, and the predictive model was developed using eXtreme Gradient Boosting (XGBoost). The model performance was evaluated using the area under the receiver operating characteristic curve (AUC).<br /><br />Results indicated that 63.1% of patients achieved a major pathological response (MPR). Over a median follow-up period of 17 months, patients achieving MPR had a significantly lower hazard ratio for event-free survival than non-MPR patients (HR 0.36, 95% CI: 0.20 - 0.65, P < 0.01). The AI model selected 4 pre-treatment features, 30 post-treatment features, and 11 delta features using LASSO regression. The AUC values for the delta model were 0.86 and 0.87 for internal and external validations, respectively.<br /><br />The study concluded that an AI-based radiomics approach could accurately predict pathological responses to neoadjuvant therapy, thereby facilitating optimal surgical timing. This could potentially improve clinical outcomes by identifying patients likely to respond well and benefit from earlier surgical intervention.

Asset Subtitle

Yang Xia

Meta Tag

Speaker Yang Xia

Topic Local-Regional NSCLC

Keywords

artificial intelligence

pathological response

neoadjuvant chemoimmunotherapy

Non-small Cell Lung Cancer

radiomics

CT scans

XGBoost

Mann-Whitney U test

LASSO regression

predictive model