Onc.AI develops AI models that analyze CT scans to improve cancer treatment decisions and optimize pharmaceutical clinical trials. Their FDA-cleared and breakthrough-designated models assist oncologists in therapy selection and identifying high-risk patients. Onc.AI offers these same regulatory-grade models to pharmaceutical companies to support data-driven decisions and increase the likelihood of success in pivotal trials. Their proprietary deep learning models, validated by real-world data and partnerships with major pharma companies, outperform conventional imaging metrics in predicting critical endpoints such as overall survival.
What is the problem?
In the clinic, oncologists often face uncertainty when selecting the right treatment path. Onc.AI’s first product, currently under FDA review, supports the PDL1-High, mut- mNSCLC patient population and helps oncologists decide between pembro and pembro+chemo using a patient's baseline CT scan. Their second product, an FDA-breakthrough designated AI model, leverages baseline and first follow-up CT scans to identify poor prognosis patients who may benefit from a change in treatment strategy, increased surveillance, and patient-oncologist discussions about care goals. In drug development, oncology trials face high attrition with 70% of Phase II and 50% of Phase III studies failing, highlighting the need for better tools.
What is their solution?
Onc.AI has developed a suite of proprietary deep learning models that automatically extract predictive and prognostic features from routinely collected diagnostic CT scans. These features are fitted to a survival model (Cox proportional hazard) to predict key oncology outcomes, including OS, PFS, and TTP. Onc.AI deploys its models through a cloud-hosted solution and offers a visual analytics interface, enabling customers to explore patient data stratified by Onc.AI scores. With their breakthrough AI model outperforming RECIST and tumor volumetrics in predicting overall survival, Onc.AI's imaging biomarkers offer a more powerful foundation for clinical development. These models can enhance trial design and decision-making across early and late phase trials by supporting patient selection, imaging covariate balancing, and futility analyses based on a more sensitive early response assessment.
