Owkin AI for identifying breast, colorectal cancer types score EU approval

In recent years, artificial intelligence developers have changed the face of medicine with algorithms capable of detecting the earliest signs of cancer. Owkin, meanwhile, is taking a slightly different approach with its own cancer-spotting AI.

Rather than training its technology to parse out microscopic indicators of disease, to aid in making an initial diagnosis, Owkin’s AI programs are designed to be used after that diagnosis to help distinguish between different forms of cancer and predict its progression.

Two of those AI models—one to predict if breast cancer patients will relapse after treatment, and another that can determine whether cases of colorectal cancer contain microsatellite instability—have now received CE mark approval, Owkin announced Friday, allowing them to be used on cancer patients across Europe.

The Owkin Dx RlapsRisk BC uses AI to analyze the digital pathology slides of patients with primary invasive breast cancer. Trained using both retrospective clinical data and previously captured pathology slides, the AI is meant to churn out a score denoting the risk that a patient will relapse within a few years of initial treatment—which happens in approximately 10% of patients, per Owkin, and dramatically decreases survival rates.

The program returns a risk score within just about 15 minutes and offers a much lower cost per analysis than the complex molecular and genetic testing typically required to predict a cancer relapse.

Study results presented last fall showed that the deep learning AI outperformed standard prediction models in estimating breast cancer patients’ five-year survival rates. At the time, the researchers shared their hopes that the tool would help reduce chemotherapy treatments—by cutting down on unnecessary treatments for low-risk patients and encouraging high-risk patients to seek potentially more effective alternatives.

“AI-based digital pathology diagnostics could help us to provide a comprehensive analysis of each tumor from just one representative standard-stained tumor slide, in a complementary process to the pathologist’s diagnosis. This would democratize access to precision medicine, unlocking a new era of treatment for patients across the world,” Magali Lacroix-Triki, M.D., Ph.D., a breast pathologist at Paris’ Institut Gustave Roussy, said in Friday’s announcement.

The other newly approved tool, Owkin Dx MSIntuit CRC, focuses on colorectal cancer. It also uses digital pathology slides to work its AI magic, but does with an aim of distinguishing between the two primary types of colorectal tumors.

The less common of the two types is microsatellite-instable colorectal cancer, which typically responds well to immunotherapy treatments and has been linked to better outcomes. Around 85% of tumors are microsatellite-stable, however, and can’t be effectively treated with an immunotherapy approach.

The MSIntuit AI can identify microsatellite-stable tumors using only standard histology slides, potentially speeding up the time it takes for colorectal cancer patients to be put on a treatment regimen that’ll cater to their specific form of the disease.