This is the first time in the world that such a tool has been clinically validated.
Artificial intelligence is not just for generate images a you text, it is also increasingly useful in the field of medicine. A study conducted by the Institut Curie recently demonstrated that an algorithm was able to detect the severity of tumors for several types of breast cancer.
2.2 million new cases each year
As the Institut Curie points out in a statement, breast cancer is most common in women with more than 2.2 million new cases each year. It is therefore necessary to obtain both faster and more accurate diagnoses to improve the chances of survival for patients.
It is for this purpose that the algorithm was created, to assist doctors and avoid diagnostic errors: “Due to the increase in the overall incidence of breast cancer and the decrease in the number of pathologists, the workload placed on pathology departments has increased significantly, explains the Institut Curie. There is therefore a growing need for automated solutions and decision support tools. »
Over 50 identifiable breast features
The company Ibex Medical Analytics, which specializes in AI cancer diagnosis, has therefore created Galen Breast to diagnose breast biopsies. Using deep learning methods, the algorithm was able to train on hundreds of thousands of images to identify more than 50 specific breast characteristics.
During the study, artificial intelligence analyzed more than 400 biopsies and the results were compared to the diagnoses made by two specialist breast pathologists. The results showed that the AI was very precise, accurately distinguishing several tumor types, including rare types. ” I had the pleasure of participating in the study and validation of new innovations that will reshape our profession for years to come, said Stuart Schnitt, chief of breast oncology pathology at the Dana-Farber/Brigham and Women’s Cancer Center and co-author of the study. I look forward to seeing more AI applications mainstream into routine clinical use as they demonstrate clinical validity. »