Artificial intelligence software was developed that can detect Kovid-19 in 50 seconds


With the artificial intelligence software developed by Istanbul University Cerrahpaşa (İÜC) Cerrahpaşa Faculty of Medicine, Sabancı University and TÜBİTAK National Metrology Institute for use in thorax computed tomography (CT), the new type of coronavirus (Kovid-19) involvement in the lung can be detected with 95 percent accuracy in 50 seconds.

IÜC Cerrahpaşa Faculty of Medicine, Head of Radiology Department. Dr. Osman Kızılkılıç reminded that Kovid-19 patients have a PCR test, computed tomography and chest radiography, which allow rapid diagnosis to take isolation measures, start drug treatment, and make hospital admissions if necessary.

Kızılkılıç stated that computed tomography is especially preferred because it is an examination method that shows much lighter attitudes, and therefore, they think how to make this method faster or how to prioritize patients who are positive in their daily work flow.

In this context, Prof. Dr. Berrin Yanıkoğlu explained that they contacted the team under the leadership of Yeşilyurt, they evaluated the images of the patients together, they read them by radiology and summarized the radiological findings in these patients, and that Sabancı University supported the computer and machine learning side.

Prof. Dr. Kızılkılıç, “With the developed software, we have succeeded in developing a system that informs us that there are positive findings of Kovid in the patient at the moment it was taken, gives a prior report to the doctor, and alerts the doctor, with a 95 percent accuracy of around 50 seconds. gave the information.

Stating that 8 people from Cerrahpaşa and 5 people from Sabancı University contributed to the project, Kızılkılıç said, “The main purpose here is to diagnose the patient quickly, to make the correct diagnosis, to make the correct diagnosis on the basis of the patient, in the form of a customized diagnosis. Therefore, these software can be further developed from now on. can be used for lung cancer screening, nodule screening, mammograms, breast cancers. What is trying to do with all these software is to give patients the correct diagnosis faster. ” he spoke.

Kızılkılıç stated that after the developed software became a prototype, it could become a commercial software or a project that could be presented to the common use of humanity.

Stating that the clinical studies of the project are continuing, Kızılkılıç said, “We are currently continuing the medical processes, but now we are using it because it has achieved 95 percent success in the clinic.” said.

It is aimed to bring the accuracy of the project closer to 100 percent.

Sabancı University Computer Science and Engineering Program Faculty Member Prof. Dr. Berrin Yanıkoğlu Yeşilyurt pointed out that at the end of March 2020, the Istanbul Development Agency has a call for investigations that can be done against Kovid-19.

Prof. Dr. He stated that Yeşilyurt, İÜC and TÜBİTAK National Metrology Institute started to work together after they came up with a common idea and that they developed this system in a period of 8-9 months.

Stating that they will continue to develop the software, Yesilyurt said, “The current seconds, that is, working under 1 minute is already very enough for us. The patient does not even leave the room during that time. Therefore, it is not accelerated, but how much closer we can bring it to 100 percent, we can try to increase the accuracy.” he spoke.

Saying that it is very enjoyable to do something to help doctors, Yeşilyurt also thanked the students who contributed to the project.

Offers great advantage in early isolation

İÜC Cerrahpaşa Faculty of Medicine Department of Radiology Lecturer Assoc. Dr. Fatih Gülşen explained that patients who applied to the emergency department with respiratory distress and had suspected Kovid-19 were asked to examine the lungs, but in hospitals with a high number of patients, tomography images were examined by radiologists and the process of informing the physician in the emergency room could take a few hours.

Stating that they started out with the desire to find a solution to this problem, Gülşen gave the following information about the working model of the developed artificial intelligence software:

“Immediately after the tomography of the patient with suspected Kovid-19 is taken, the tomographic images of the patient are transmitted to the artificial intelligence software, after the software scans the images, Kovid-19 findings are sent to the automation system of our hospital. The entire process is completed in less than 1 minute, in an average of 45 to 50 seconds. Thus, with the warning of the artificial intelligence system, the patient is rapidly isolated while still in the tomography shooting room. It is possible to evaluate the tomographies of the patients who are given a Kovid warning by the system, to inform the emergency department physician and to start the treatment of these patients earlier. r brings us an advantage. “

Assoc. Dr. Gülşen added that they think that the artificial intelligence system, which they aim to continuously develop in accordance with the computed tomography devices and medical conditions in Turkey, will be very useful especially in hospitals with high patient density, so they wish it to become widespread throughout the country.