Site Editor
Gregory J. Riely, MD, PhD
Gregory J. Riely, MD, PhD
Posted: Monday, February 3, 2025
Artificial intelligence (AI)-based deep learning models may prove useful in improving clinical outcomes in the treatment of advanced non–small cell lung cancer (NSCLC), guiding immune checkpoint inhibitor (ICI) treatment, and better identifying patients who could benefit from it, according to research published in JAMA Oncology. David J. Kwiatkowski, MD, PhD, of Brigham and Women’s Hospital, Harvard Medical School, Boston, and colleagues developed and independently validated a deep learning−based response stratification model for predicting ICI treatment outcomes in patients with advanced NSCLC from whole-slide hematoxylin and eosin (H&E)-stained images. Deep learning prediction scores were associated with outcomes across patient cohorts and combined deep learning and PD-L1 scores improved patient stratification.
“This is the first proof-of-concept study to devise an AI-driven model for predicting ICI response in advanced and metastatic stages of NSCLC using digital H&E pathology images,” the investigators stated. “The deep learning model has the capability to predict ICI responses directly from a single image of an H&E-stained slide.”
A total of 295,581 image tiles from 958 patients treated with ICI for NSCLC were included in the multicenter cohort study. Participants were grouped into a developmental cohort drawn from one participating center in the United States and a validation cohort drawn from three facilities in the European Union.
The researchers reported that the objective response rate to ICI was 26% in the developmental cohort and 28% in the validation cohort. In multivariable analysis, the deep learning model’s score was found to be an independent predictor of ICI response in the validation cohort for both progression-free (hazard ratio [HR] = 0.56; 95% confidence interval [CI] = 0.42–0.76; P < .001) and overall survival (HR = 0.53; 95% CI = 0.39–0.73; P < .001).
In addition, the investigators found that the tuned deep learning model achieved a higher AUC than tumor mutational burden, tumor-infiltrating lymphocytes (TILs), and PD-L1 in the developmental cohort. In the validation cohort, it was found to be superior to TILs and comparable with PD-L1. Finally, in the validation cohort, combining the deep learning model with PD-L1 scores outperformed either marker alone.
Disclosure: For full disclosures of the study authors, visit jamanetwork.com.
JNCCN Spotlights
