AI Tool Achieves 91% Accuracy Assessing Lung Cancer Treatment Response Across Global Studies

AI Algorithm Delivers 91% Accuracy in Assessing Pathologic Response

An AI algorithm has achieved 91% accuracy in assessing pathologic response to neoadjuvant chemoimmunotherapy in patients with non-small cell lung cancer, demonstrating consistent performance across four international cohorts from Spain, China, and Germany1

1

. The AI tool showed at least moderate concordance with pathologist assessments, with discordant classifications occurring in fewer than 9% of cases. The study analyzed 135 patients and 1,344 haematoxylin and eosin-stained slides from primary tumours, with the AI algorithm quantifying tumour, necrosis, and stroma areas to generate weighted average percentages of residual viable tumor.

Robust Performance Metrics Across International Validation

The AI tool demonstrated exceptional performance in identifying major PR, defined as less than or equal to 10% residual viable tumor in the primary tumour. The pooled sensitivity and specificity reached 90% and 93% respectively, with an AUC score of 0.95 or higher across all four cohorts1

1

. Discordant classifications between manual and AI-based assessment occurred in only 12 of 135 cases, predominantly false negatives. The main sources of error included stromal underrepresentation and misclassification of reactive pneumocytes and multinucleated giant cells, highlighting specific areas for algorithm refinement.

I3LUNG Project Transforms Immunotherapy Prediction

The I3LUNG project enrolled 2,396 patients with advanced non-small cell lung cancer treated with immunotherapy across six centers in Italy, Germany, Greece, Israel, Spain and the United States2

2

. The international research team integrated clinical, imaging, pathology, and digital pathology data into comprehensive multimodal AI models. These models consistently outperformed all standard clinical biomarkers, including PD-L1, which currently guides treatment decisions despite well-known limitations. The AI model using clinical and blood data achieved an AUC score of 0.77, while the model incorporating clinical, blood, imaging, and pathology data reached 0.88.

Human-AI Collaboration Enhances Clinical Decision-Making

Twenty physicians reviewed 100 real patient cases, first without and then with AI support, revealing significant improvements in diagnostic accuracy. Access to the AI tool improved sensitivity for identifying responders from an AUC score of 0.72 to 0.872

2

. Physicians who were not lung cancer experts showed the greatest improvements, a finding with direct relevance to community oncology settings where thoracic oncology expertise may be limited. Inter-physician agreement rose from slight to moderate, suggesting the AI tool promotes more consistent clinical reasoning across different experience levels. "This alignment between machine and clinical logic is essential for building trust in AI-assisted decision-making," said Marina Garassino, MD, Professor of Medicine at UChicago Medicine and senior author of the I3LUNG study.

Addressing Critical Gaps in Immunotherapy Treatment

Immunotherapy has transformed lung cancer treatment, achieving long-term benefit in 20% to 30% of patients, yet most patients experience resistance to treatment2

2

. Physicians cannot reliably predict which patients will benefit from immunotherapy, leading to unnecessary toxicity and cost for non-responders. Better predictive clinical biomarkers could identify likely responders at diagnosis, enabling physicians to better tailor treatments for individual patients. The AI tool developed through the I3LUNG project addresses this critical need by providing expert-level guidance at the point of care, particularly valuable in community oncology settings.

Future Implications for Precision Oncology

The I3LUNG project is now prospectively enrolling more than 2,000 patients across the same six international centers, focusing on treatment optimization and clinical usability2

2

. As perioperative treatment strategies continue to evolve and the demand for reliable surrogate endpoints increases, AI-assisted quantification of residual viable tumor offers a practical pathway toward greater standardization. "I3LUNG establishes a new benchmark for AI in thoracic oncology," Garassino noted. "For patients, this means fewer missed opportunities for treatment benefit. For the field, it provides a rigorous, fair, and explainable framework that can serve as a global platform for the next generation of precision immunotherapy."