Owkin's AI model predicting cancer patient survival from radiology and clinical data published in European Journal of Cancer

For the past three years, Owkin has been leading PULS-AI – a research project to develop new biomarkers to improve the targeting of antiangiogenic drugs, a cancer treatment approach that aims to inhibit the growth of a tumor by preventing the formation of new blood vessels. Finding a prognostic marker for antiangiogenic therapy could help doctors to plan their care by identifying patients who are most likely to benefit from treatment.

Working with researchers from Gustave Roussy, a leading cancer center in Europe, we used deep learning to train a model on CT scans, ultrasound images and clinical data from more than 600 patients from 17 French treatment centers. Our goal was to develop a reliable way to predict the outcome of patients with seven different cancers treated with antiangiogenics.

Today, the results have been published in the European Journal of Cancer.

By using AI to analyze radiology and clinical data, PULS-AI successfully predicts the outcomes of cancer patients treated with antiangiogenics

PULS-AI serves as a new biomarker that can accurately predict how likely a patient is to survive under antiangiogenic treatment, a tool that could provide therapeutic decision-making assistance to clinicians.

*The PULS-AI model can accurately stratify patients into a high-risk and low-risk group, based on the Kaplan–Meier method.*

‍

PULS-AI project has led to the creation of a unique database of manually annotated radiology images, thanks to Gustave Roussy’s radiologists: 1,147 US images were annotated with lesions delineation and 632 thorax-abdomen-pelvis CTs (total of 301,975 slices) were fully-annotated with a total of 9,516 lesions.

PULS-AI is a composite, multimodal biomarker, the approach of which could now be applied to other treatment and disease areas

It has shown that AI is able to extract relevant information from radiology images and aggregate data from different modalities to build powerful prognostic tools that could improve therapeutic decisions. We also demonstrated that training a model on both imaging and clinical data yields significantly better results than single modalities.

PULS-AI shows the benefit of assessing patients’ complete tumor burden on CT scans. It emphasizes the importance of integrating AI algorithms in the radiologist’s workflow to automate and homogenize this time-consuming annotation process.

We are grateful for the financial support from the Ile De France Region.

Read the full paper in the European Journal of Cancer.