Collaboration between pharmaceutical companies and the Center for Computational Imaging and Personalized Diagnostics (CCIPD) at Case Western Reserve University has led to the development of artificial intelligence (AI) tools to benefit patients with non-small cell lung cancer (NSCLC) based on an analysis of routine tissue biopsy images, according to new research. 

This year, more than 236,000 adults in the United States will be diagnosed with lung cancer—about 82% of them with non-small cell lung cancer, according to the American Society of Clinical Oncology.

Researchers at the CCIPD used AI to identify biomarkers from biopsy images for patients with NSCLC, as well as gynecologic cancers, that help predict the response to immunotherapy and clinical outcomes, including survival.

“We have shown that the spatial interplay of features relating to the cancer nuclei and tumor-infiltrating lymphocytes drives a signal that allows us to identify which patients are going to respond to immunotherapy and which ones will not,” said Anant Madabhushi, CCIPD director and Donnell Institute Professor of Biomedical Engineering at Case Western Reserve.

Immunotherapy is expensive, and studies show that only 20-30% of patients respond to the treatment, according to the National Institutes of Health and other sources. These findings validate that the AI technologies developed by the CCIPD can help clinicians determine how best to treat patients with NSCLC and gynecologic cancers, including cervical, endometrial, and ovarian cancer, Madabhushi said.

The study, drawn from a retrospective analysis of data, also revealed new biomarker information regarding a protein called PD-L1 that helps prevent immune cells from attacking non-harmful cells in the body.

Patients with high PD-L1 often receive immunotherapy as part of their treatment for NSCLC, while patients with low PD-L1 are often not offered immunotherapy, or it’s coupled with chemotherapy.

“Our work has identified a subset of patients with low PD-L1 who respond very well to immunotherapy and may not require immunotherapy plus chemotherapy,” Madabhushi said. “This could potentially help these patients avoid the toxicity associated with chemotherapy while also having a favorable response to immunotherapy.”

The multi-site, multi-institutional study examined three common immunotherapy drugs (called checkpoint inhibitor agents) that target PD-L1: atezolizumab, nivolumab, and pembrolizumab. The AI tools consistently predicted the response and clinical outcomes for all three immunotherapies.

The study is part of broader research conducted at CCIPD to develop and apply novel AI and machine-learning approaches to diagnose and predict the therapy response for various diseases and cancers, including breast, prostate, head and neck, brain, colorectal, gynecologic, and skin.

The study coincides with Case Western Reserve recently signing a license agreement with Picture Health to commercialize AI tools to benefit patients with NSCLC and other cancers.