One of the most intriguing characteristics of the COVID-19 pandemic is that the disease acts differently from one person to another. Centro Diagnostico Italiano, a diagnostic center based in Milan, Italy, has partnered with Amazon Web Services Inc. to apply artificial intelligence and machine learning to predict patients who are likely to experience a severe form of COVID-19 and intervene more quickly with an appropriate level of care.

“We select only positive patients … and we try to evaluate from the first chest X-ray what will be the real destiny of the patient, better or worse,” said Dr. Sergio Papa (pictured), director of diagnostic imaging at the Centro Diagnostico Italiano. “And then we can also predict — try to predict, obviously — how many intensive-care beds are a necessity in that institution. We can send the therapy and adjust the therapy for the different groups of patients.”

The overall project is called Al-for-COVID and is supported by donated computing credits provided through the AWS Diagnostic Development Initiative. The study is a public-private partnership involving the Centro Diagnostico Italiano, diagnostic imaging company Bracco Imaging, and a number of Italian institutes and hospitals.

Papa spoke with Stu Miniman, host of theCUBE, SiliconANGLE Media’s livestreaming studio, during the AWS Public Sector Summit event. They discussed details of the project, how technology will be used in various parts of the study, and how the research will be useful during the current pandemic and beyond. (* Disclosure below.)

Machine learning identifies X-ray patterns

AWS technology will support research in many areas, according to Papa. First, the partnership involves developing a platform on which hospitals and institutes can store their sets of images. The idea is to make all the images available in one place to make working with them easier.

In addition, AWS technology will apply machine learning to analyze chest X-ray images of diagnosed patients who enter hospitals.

“And, third, I think could be … generating the structure of the reports for this patient and, moreover, the identification of patterns, different patterns, that we can find inside the images. This concerns the radiomics theory,” Papa explained.

In medicine, radiomics is a method that extracts a large number of features from radiographic medical images using algorithms. These so-called radiomic features have the potential to discover diseases that are not appreciated by the naked eye.

“If I check an image in a CT scan or an X-ray, I can see inside that image. The maximum that I can see is 10, 15 to the maximum different patterns,” Papa said. “If a machine system examines the same image, it can reach out hundreds of patterns that I cannot see.”

After identifying these features, the system will compare them with a very huge database already built and try to understand which patterns are linked to different outcomes. “So we can say, ‘Okay, this image has 20, 30, 100 patterns that suggest to us the destiny of the lesion will be one specific, while another lesion — that for me is exactly the same with my eyes — bu systems will tell us … [its] destiny is really different,” Papa stated.

The partnership with AWS will also allow the Centro Diagnostico Italiano to share its research with other scientific centers and, in the end, share the results with the scientific community around the world.

“Our study is focused on pneumonia from COVID-19, but the methodology can be applied in every kind of interstitial pneumonia,” Papa explained. “Usually in radiomics, we used to study the single lesions or little areas. This is one of the first, very first important studies where the segmentation is dedicated to the whole organ — I mean all the lung, both lungs.”

Watch the complete video interview below, and be sure to check out more of SiliconANGLE’s and theCUBE’s coverage of the AWS Public Sector Summit event. (* Disclosure: TheCUBE is a paid media partner for the AWS Public Sector Summit Online event. Neither Amazon Web Services Inc., the sponsor for theCUBE’s event coverage, nor other sponsors have editorial control over content on theCUBE or SiliconANGLE.)

Photo: SiliconANGLE