FDA Grants Breakthrough Device Designation to Serial CTRS AI Model for NSCLC

The FDA has granted breakthrough device designation to Serial CTRS, an artificial intelligence (AI)–based prognostic tool intended to stratify patients with non–small cell lung cancer (NSCLC) into high- or low-risk mortality categories.1

Serial CTRS utilizes a deep-learning model to classify patients with NSCLC into different risk categories. The prognostic tool is part of a pipeline of AI imaging models being developed by Onc.AI to automate risk prognosis and optimize treatment decision-making for patients with NSCLC in order to enable more precise, personalized care.

“We are honored to be awarded breakthrough device designation for our Serial CTRS AI model,” Akshay Nanduri, chief executive officer of Onc.AI, stated in a news release. “Onc.AI aims to equip oncologists with vital, automated prognostic insights using routinely collected diagnostic imaging scans and ultimately improve treatment strategy and provide risk stratification throughout [the] journey [for a patient with cancer].”

Data presented at the 2024 SITC Annual Meeting from a multi-institutional study demonstrated that Serial CTRS generated improved overall survival (OS) predictions compared with standard assessment tools in patients with NSCLC receiving immunotherapy.2 During the study, using Serial CTRS on CT scans at baseline and at 3 months of follow-up was found to accurately predict long-term outcomes after only a few cycles of treatment.

Findings showed that C-index for predicting OS was improved with Serial CTRS (0.734) compared with RECIST (0.631) and tumor volume measurement changes (0.679). Serial CTRS was also found to be a significant predictor of OS after adjusting for other factors, such as tumor volume change, PD-L1 tumor proportion score, age, sex, and line of therapy.

In patients with stable disease, Serial CTRS produced a 12-month area under the receiver operating characteristic curve of 0.74 (95% CI, 0.65-0.82) compared with 0.62 (95% CI, 0.52-0.72) for tumor volume change.

To develop the model, researchers used a real-world dataset comprised of patients with advanced NSCLC who were treated with immune checkpoint inhibitors.3 A pipeline of image quality control, preprocessing, deep-learning feature extraction, and a survival model were generated based on serial CT scans.

The AI model was then validated using additional patients from the real-world dataset, where hazard ratios for OS were compared with tumor volume change stemming from manual volumetric segmentations and RECIST 1.1 categories of response. Serial CTRS and volume change were categorized as high, medium, or low response.

“As longstanding partners of Onc.AI, we are thrilled to see the application of Flatiron’s high-quality, curated real-world data in the development and validation of regulatory-grade AI models for clinical use,” Jacqueline Law, vice president of Corporate Strategy at Flatiron Health, stated in a news release.1 “We look forward to supporting Onc.AI’s efforts collaborating with the FDA and achieving additional milestones together.”

In a news release, Onc.AI also noted that Serial CTRS could play a role in oncology drug development programs, including trial design and clinical development decisions for novel therapeutics.

“As part of our ongoing data and clinical collaboration with Onc.AI, we are excited to be evaluating Serial CTRS. Having been involved in product definition and evaluating results throughout the evolution of this product, I look forward to seeing this breakthrough technology enter the clinic and impact early phase trials and clinical development,” Dwight Owen, MD, MS, associate professor of medicine and head of Thoracic Oncology at The James Cancer Center at Ohio State University, added in a news release.

References

1. Onc.AI awarded FDA breakthrough device designation for Serial CT Response Score deep learning model. News release. Onc.AI. February 6, 2025. Accessed February 7, 2025. https://www.businesswire.com/news/home/20250206095197/en/Onc.AI-Awarded-FDA-Breakthrough-Device-Designation-for-Serial-CT-Response-Score-Deep-Learning-Model
2. Onc.AI Serial Imaging Response Score outperforms traditional methods for early assessment of NSCLC immunotherapy outcomes. News release. Onc.AI. November 6, 2024. Accessed February 7, 2025. https://www.businesswire.com/news/home/20241106220858/en/Onc.AI-Serial-Imaging-Response-Score-Outperforms-Traditional-Methods-for-Early-Assessment-of-NSCLC-Immunotherapy-Outcomes
3. Sako C, Schmidt TG, Patel AA, et al. Deep learning serial CT response score predicts overall survival in advanced NSCLC treated with PD-(L)1 immune checkpoint inhibitors. J Immunother Cancer. 2024;12(suppl 2):1237. doi:10.1136/jitc-2024-SITC2024.1237