As Barbara Ban/Novac writes on the 3rd of April, 2020, any help during this unprecedented pandemic is more than welcome. That is why young computer scientists from Pula – Kristijan Saric and Matija Plastic decided to focus their knowledge on creating an application that could help medical staff recognise the presence of coronavirus in the future. Meet Exactbyte.
”We’re a company from Pula that is primarily engaged in artificial intelligence and machine learning, and the development of applications using these technologies. We’d like to use this knowledge and experience to help with the COVID-19 virus situation. We have developed a machine learning model that could help doctors diagnose COVID-19 patients.
The model was trained to recognise whether or not lungs on lung x-rays are healthy, have pneumonia or if there is COVID-19 present,” explained 33-year-old Kristijan Saric, who opened Exactbyte three years ago and hired 28-year-old Matija and another student. He says that when it comes to the preparation of this application, which could one day help medicine, he started before the epidemic began in Croatia, by taking numerous photos of lung x-rays from coronavirus positive patients, and then those of healthy lungs, and the lungs of those with inflammation.
”I took x-rays I found on the internet. These are about seventy images of the lungs of coronavirus patients. I also used 500 images of healthy lungs and nearly 4500 images of the lungs of people with pneumonia. The programme we developed uses these images for testing and checks their status based on image recognition,” explains Exactbyte’s founder.
However, for the application to come to life and function properly, Kristijan says he needs a lot more pictures of lungs infected with coronavirus, which is why he is seeking cooperation with healthcare institutions. The end goal is to implement those images into a web application that would allow access to doctors and hospitals.
”They could send an x-ray and get feedback from the system on whether it is a healthy lung, a lung with pneumonia or a lung with COVID-19, and mark the portion of the image on which the system reached a conclusion so that they can verify it,” Kristijan explained.
He had already put the idea out, he says, to the Pula General Hospital, with which he had already started collaborating on a project before the epidemic broke out, namely the classification of breast cancer. Otherwise, most of the work Exactbyte does is for an American company.
Make sure to follow Made in Croatia for more Croatian innovation.
