Computer Vision News - April 2020

2 Summary RSIP Vision's projects 2 findings in the chest and it can make a quantitative as well as qualitative output and correlate this finding to what phase the patient is in now. This scan makes our life really easy and can save a lot of time.” In order to achieve this knowledge, the machine would need to be run many thousands of times to train the network . It is theoretically possible to use CT scans obtained in China, North Korea and Japan, where the first cases were discovered. “Chest CT for people suspected of having COVID-19 doesn’t need any special protocol,” Rabeeh points out “and the disease can be detected using any scan that we can find. Low-dose, high-resolution, with contrast or without contrast.” Since the goal is to detect the presence of the disease and (in positive cases) Mount Sinai Hospital at what stage it is, a deep learning classification process is recommended. Once the available scans have been correctly annotated by experts, the classification is done using one of the many available deep neural networks. Quantitative findings can be obtained through a process of segmentation with deep learning that will give precise results, far beyond the ones detected by the human eye . Finally, how would Rabeeh assess RSIP Vision’s capability to provide this kind of solution to accurately detect coronavirus on CT scans? “From my experience with the company, it’s really high. I think that RSIP Vision’s engineers are capable and they have the tools and expertise at their disposal to do it right now!”