Are coronavirus tests history?
With the 'Artificial Intelligence System', the corona virus can be detected 100 percent.
Ağrı İbrahim Çeçen University Computer Department, Assoc. Dr. Aytürk Keleş, his wife Assoc. Dr. Together with Ali Keleş and his sons, Mustafa Berk Keleş, who is a student of Istanbul Aydın University Software Engineering, he developed an artificial intelligence system that diagnoses the corona virus 100 percent.
They stated that they have previously carried out many studies on the diagnosis of diseases with artificial intelligence and that they started this work to contribute to the fight against the epidemic during the period when they closed home due to the pandemic. Associate Professor. Aytürk Keleş said in his statement that while his son Mustafa Berk Keleş was still a student, he worked as an Artificial Intelligence expert in two different international technology companies and that he was included in this study with his expertise in the field.
"Our work is a rare work in the world that will contribute to this field"
Stating that his work is rare worldwide, Assoc. Aytürk Keleş said that they worked closely with her husband and child during the pandemic process as a team and said, “We started this work by thinking about what we can do together, how we can contribute to the fight against this disease during the period when we closed home due to the pandemic. Using deep learning from artificial intelligence technologies, we have developed intelligent inference mechanisms that will turn cheap, portable but low-sensitivity X-ray devices into powerful diagnostic tools for the diagnosis of Covid-19. Our study is a rare study in this field in the world. Because all the work done by experts to detect Covid-19 so far has used ready-made networks trained on general images. For example VGG19, MobileNet, DenseNet and SqueezeNet. However, medical images are different in many ways. Therefore, developing networks from scratch with Covid-19 images is very important in order to be used in diagnostic tasks. However, developing deep learning networks from scratch is difficult as it requires both experience and serious expertise in the field. Using X-ray images, we developed two different architectures from scratch, and deep learning networks that we call COV19-CNNet and COV19-ResNet. These networks are in the form of powerful inference mechanisms that can also be integrated into X-ray devices. These mechanisms can diagnose normal and Covid-19 cases with 100 percent accuracy and viral pneumonia with 97 percent accuracy.
"We can detect corona virus in less than a second from cough or breath"
Emphasizing that their work has been published in many international journals with a high academic level, Assoc. Aytürk Keleş said, "We started our study when the disease first appeared in our country. We finished it in three months. We waited for the publication phase of our study for five months to show its international validity.