AI Tool Predicts Chemotherapy Response in Small Cell Lung Cancer

Small cell lung cancer often presents at an extensive stage and progresses rapidly, leaving little time to tailor first-line therapy. Clinicians currently lack biomarkers to guide which patients will benefit from platinum-based chemotherapy, leading to delayed switches and missed trial opportunities. Researchers now report an artificial intelligence (AI) pathology tool that predicts response before treatment begins and without additional biopsies. The approach aims to help direct patients to effective regimens sooner and clarify prognosis.

The AI tool, PhenopyCell, functions as a computational biomarker that assesses standard pathology slides and other routinely collected clinical data. It predicts whether a patient with extensive-stage small cell lung cancer (SCLC) will respond to platinum-based chemotherapy ahead of treatment initiation. The accuracy of the method has been verified by the three institutions that collaborated on its evaluation.

PhenopyCell operates by characterizing immune–tumor architecture within diagnostic biopsy tissue. It integrates visual patterns from pathology slides with information drawn from patients’ medical records to determine how these features correspond to outcomes. In doing so, it addresses the absence of validated SCLC biomarkers that can guide first-line decision-making.

In a retrospective study, investigators applied PhenopyCell to slides from 281 patients with SCLC treated at Roswell Park Comprehensive Cancer Center, Winship Cancer Institute of Emory University, and University Hospitals Cleveland Medical Center. The tool predicted response to platinum-based chemotherapy from the diagnostic biopsy tissue alone, before therapy began. Predictions were compared with observed outcomes and demonstrated greater accuracy than manual review.

The analysis revealed that tumors associated with better outcomes contained more immune cells organized in groups around tumor clusters, suggesting a more effective immune response. Poorer outcomes were linked to fewer immune cells, which appeared in disorganized groups farther from tumor cells. These spatial patterns were visible only with the AI-based pathology approach. Study findings were published in npj Precision Oncology on March 17, 2026.

“Every patient with small cell lung cancer already has a pathology slide from their diagnostic biopsy,” said Prantesh Jain, MD, FACP, thoracic oncologist at Roswell Park Comprehensive Cancer Center. “This system works from that existing slide. There’s no need for additional procedures or tissue collection, and no added cost. In a disease where survival is measured in months and re-biopsy is rarely possible, this has the potential to become a uniquely powerful tool.”

Related Links
Roswell Park Comprehensive Cancer Center.