Abstract P93: Multi-Cancer Screening Using AI-Enhanced Canine Olfactory Detection: Interim Results from a Multi-Center Breath Analysis Study in India
Sanjeev Kulgod, Basavaraj R Patil, Shashidhar K, Rakesh S Ramesh, Somashekhar S P, Swaratika Majumdar, Sahana Shanbhag, Akash Kulgod, Itamar BitanAbstract
Background:
Early cancer detection remains a critical challenge in global healthcare. Breath is an ideal first-tier screening matrix: non-invasive, sampleable outside clinics, and rich in volatile organic compound (VOC) signatures generated by tumor metabolism. We developed BreathEasy, a digital workflow that converts canine olfaction into quantifiable data streams for multi-cancer screening.
Methods:
This assessor-blinded, multi-center case-control study was conducted across six hospitals from October 2024 to June 2025. Each breath sample was presented in a sensor-instrumented arena; we captured binary dog indications, dwell-time heat maps from sensors and cameras, behavioral inference from computer vision models, and dog-specific historical performance. A Bayesian voting model fused these signals with sample covariates to generate screening calls and uncertainty scores. Operating thresholds were set in 1,250 participants and locked. We present results from the independent validation cohort of 1,250 participants (250 biopsy-confirmed cancers, 1,000 controls).
Results:
The Bayesian ensemble model demonstrated 93% sensitivity and 88% specificity for multi-cancer detection. Performance remained consistent across tumor types with comparable sensitivity in stage I-II disease. Breath samples remained stable for 180 days at -20°C, demonstrating operational feasibility. The system delivered calibrated, high-sensitivity probabilities with uncertainty quantification, enabling downstream triage decisions.
Conclusions:
This study establishes the clinical validity of digitally-augmented canine olfactory detection for multi-cancer screening. BreathEasy provides scalable, low-cost (<US$1 consumables) risk stratification that can decouple expensive multi-cancer early detection assays from population prevalence. This blinded dataset is one of the largest to date studies to demonstrate the potential for population-level breath-based multi-cancer screening.
Clinical Trial Registration:
Clinical Trials Registry of India.
Citation Format:
Sanjeev Kulgod, Basavaraj R Patil, Shashidhar K, Rakesh S Ramesh, Somashekhar S P, Swaratika Majumdar, Sahana Shanbhag, Akash Kulgod, Itamar Bitan. Multi-Cancer Screening Using AI-Enhanced Canine Olfactory Detection: Interim Results from a Multi-Center Breath Analysis Study in India [abstract]. In: Proceedings of Frontiers in Cancer Science 2025; 2025 Nov 5-7; Singapore. Philadelphia (PA): AACR; Cancer Res 2026;86(13_Suppl):Abstract nr P93.