In video, above, IBM Fellow Tanveer Syeda-Mahmood, principle investigator applies Watson’s machine learning capabilities to medical image analysis, ultimately providing evidence-backed diagnoses for clinicians to review.
Watson has identified cases of heat disease previously undiagnosed by human inspection of ultra-sonic imaging, clinical tests and case notes, according to IBM at the Conference on Medical Image Computer Assisted Intervention (MICCAI). As a result, IBM is offering five new Watson modules to doctors aimed at improving the accuracy of diagnoses for stroke, mammography, radiology and cancer.
So far, IBM has proved the concept by using Watson to inspect diagnoses, echocardiograms and ultra-sonic images of past patints. By training on a large database of correctly diagnosed cases—including unstructured data such as doctor’s notes—Watson was able to automatically spot 25% more heart disease from a dataset sample of cases it had never before seen, according to IBM Fellow Dr. Tanveer Syeda-Mahmood.
The study was led by Syeda-Mahmood, whose father was misdiagnosed, subsequently receiving the wrong medication which put him into a coma. Read all the details in her book chapter Identifying Patients at Risk for Aortic Stenosis Through Learning from Multimodal Data.
Watson’s new Health Imaging and Merge tools include MedyMatch (to help emergency room physicians quickly spot stroke victims), Marktation (that automatically spots abnormalities in mammograms), Clinical Integration Module (to pinpoint abnormalities in medical images) and the Lesion Segmentation and Tracking Module (to help radiologists identify cancer).
IBM
ibm.com
