Retrospective study finds multimodal AI improves immunotherapy outcome prediction in advanced lung cancer

Artificial intelligence models developed through the international I3LUNG project outperformed standard clinical biomarkers at predicting outcomes in patients with advanced non-small cell lung cancer receiving immunotherapy. The retrospective study, published in Nature Medicine, found that a model integrating clinical, blood, imaging and digital pathology data achieved an area under the receiver operating characteristic curve of 0.88 for survival-outcome prediction. A clinical-and-blood model achieved an AUC of 0.77, while AI-assisted physician review improved identification of likely treatment responders within the study’s 100-case evaluation.

The research included 2,396 patients with advanced non-small cell lung cancer who received immunotherapy at six centers in Italy, Germany, Greece, Israel, Spain and the United States. During the project’s first, retrospective phase, investigators assembled clinical information, blood data, medical images, pathology information and genomic data in a common database. They then developed and tested two families of AI models intended to predict treatment response and survival. One model used clinical and blood data, while the other added imaging and digital pathology information.

Model discrimination was evaluated with the area under the receiver operating characteristic curve, or AUC, a machine-learning measure of how accurately a model distinguishes relevant classes. The clinical-and-blood model reached an AUC of 0.77 for survival-outcome prediction. The multimodal model combining clinical, blood, imaging and digital pathology data reached 0.88, a level described in the research as excellent. Both models consistently outperformed the standard clinical biomarkers evaluated by the investigators, although the reported results concern retrospective prediction and do not demonstrate that treatment selected by the tool improves survival.

The study also examined whether the AI tool could support clinical judgment. As part of the I3LUNG study, 20 physicians—10 lung cancer experts and 10 physicians from other specialties—reviewed 100 real patient cases, first without and then with AI support. Access to the tool increased the sensitivity of responder identification, with the corresponding AUC rising from 0.72 to 0.87. Inter-physician agreement increased from slight to moderate, indicating more consistent assessments across participants with different levels of expertise. Physicians outside lung cancer specialization showed the greatest improvements, a finding relevant to clinical settings where thoracic oncology expertise may be less readily available.

The findings do not yet establish the tool’s effectiveness in routine prospective care because the reported study covered the project’s retrospective phase, and the physician evaluation was limited to 100 cases. I3LUNG is now prospectively enrolling more than 2,000 patients at the same six international centers, with a focus on treatment optimization and clinical usability. The researchers said AI evaluation should consider both predictive performance and practical use in clinical workflows. The study was supported by the European Union’s Horizon 2020 research and innovation program.

Subject of Research: Multimodal AI for predicting immunotherapy outcomes in advanced non-small cell lung cancer

Article Title: AI tool successfully predicts outcomes of immunotherapy in lung cancer

Article References: AI tool successfully predicts outcomes of immunotherapy in lung cancer. Original publication (Bibliographic details unavailable.)

Image Credits: AI Generated

DOI: Not provided

Keywords: artificial intelligence, lung cancer, immunotherapy, non-small cell lung cancer, survival prediction, clinical biomarkers, medical decision support, digital pathology, medical imaging, multimodal models, tool, successfully