Bulletin No. 1, 2021

COVID Diagnosis Made Better, Faster and Safer An AI system can now rapidly and accurately detect COVID-19 infections in chest CT images using the federated learning technique. The model is trained to recognize COVID-19 lesions independently at multiple hospitals with their own stock of CT scans. Their insights are then gathered and contribute to a central model, which returns correct diagnosis 95% of the time. Throughout the process, the training images remain with individual hospitals without having to go into the same pool and potentially leading to privacy breaches. A doctor will need five to 10 minutes to study a chest CT scan. Assessments are made in qualitative terms. In just around 40 milliseconds, the model, working on a convolutional neural network (CNN), delivers its diagnosis with indications of the lesions’ locations and severity in quantitative terms. OPERATIONAL STAGE RAW CT IMAGE AI-PROCESSED CT IMAGE Dr. Tiffany So and Prof. Simon Yu Department of Imaging and Interventional Radiology Prof. Dou Qi and Prof. Heng Pheng-ann Department of Computer Science and Engineering 29 THE NEW GOSPEL ACCORDING TO A.I.

RkJQdWJsaXNoZXIy NDE2NjYz