AI more accurately identifies patients with advanced lung cancer that respond to immunotherapy and helps doctors select treatments

Newswise — Treatment planning for lung cancer can often be complex due to variations in assessing immune biomarkers. In a new study, Yale Cancer Center researchers at Yale School of Medicine used artificial intelligence (AI) tools and digital pathology to improve the accuracy of this process.

Researchers compared AI-powered digital scoring with traditional manual scoring of the PD-L1 immune biomarker to determine if a new immunotherapy treatment, atezolizumab, could benefit patients with advanced non-small cell lung cancer. PD–L1 expression is considered the best biomarker to predict responsiveness to immune-checkpoint inhibitors.

Roy S. Herbst, lead study author and deputy director of Yale Cancer Center, will present the new findings at the World Conference on Lung Cancer in Singapore on Sept. 11.

“Our study suggests that artificial intelligence has the ability to improve the identification of PD-L1 positive patients by providing a predictive accuracy that was better than manual scoring,” said Herbst, who is also the assistant dean of translational research at Yale School of Medicine. “The research underscores the potential of digital pathology and AI tools in enhancing PD-L1 scoring accuracy for both clinical practice and clinical trials.”

To conduct this study, researchers used data from the phase III trial IMpower 110, which tested the effectiveness of atezolizumab compared to chemotherapy as a first-line treatment for advanced non-small cell lung cancer (NSCLC). Using both manual and AI-powered tumor cell scoring, researchers found that the AI model was able to identify more patients as PD-L1 positive compared to the conventional manual scoring.

The study also demonstrated that both manual and digital scoring methods were equally adept at predicting patient outcomes, including overall survival and progression-free survival. The AI model also helped conclude that among patients with squamous histology (a specific subtype of NSCLC), the presence of PD-L1+ lymphocytes correlated with improved progression-free survival when treatment included atezolizumab.

“The insights gained with AI and digital scoring could make diagnosing and choosing the right treatment easier,” said Herbst. “Our data shows that this AI technology can help refine strategies for treating advanced non-small cell lung cancer.”