If the world had not faced covid-19, Daniel Kvak and his wife Karolina might have tackled bone age prediction. Instead, they focused on the lungs, using artificial intelligence to develop a system that can tell in seconds from an X-ray whether a patient is healthy or has covid, pneumonia or any other lung finding, including a tumour.
He is a PhD student at Masaryk University in Brno, working on machine learning, she is a forensic anthropologist. That's not the beginning of the advert, but the beginning of the story of the Carebot startup. There is still a third to the party, Matěj Misař, formerly a journalist specializing in healthcare, now Carebot's CEO. The word was given when all three of them realized that this idea has a much bigger scope and that it will not be "just" about covid, but that it can completely help the whole healthcare sector.
"Imagine that every radiologist and every doctor has a system on their computer through which they look at X-ray images. We put our model directly into their system. When he receives a photo of a patient's lungs, he clicks a button and Carebot immediately recommends what the patient has on his lungs and where it is," explains Matěj Misař.
Carebot uses state-of-the-art machine learning and computer vision technologies, or works on the basis of deep neural networks. "We present the software, the neural network, with images of the lungs and at the same time annotations, i.e. what the patient's findings are. This makes Carebot able to learn to recognise individual findings based on this knowledge alone. It has been trained on hundreds of thousands of images of women, men and children that have been annotated twice, i.e. evaluated twice by different doctors," explains Daniel Kvak.
We trained our neural network on a large dataset of individual findings, age groups, geographic locations or gender, taking into account standing or lying patients, but most importantly also the possibilities of how to scan and how to scan. This is also crucial for other countries where we encounter images of varying quality. Carebot can handle these as well.
Carebot's ambition is certainly not to replace the doctor - he will always be there to diagnose and treat patients. And doctors realise this, they perceive Carebot positively. As Daniel Kvak says, radiologists are probably the most open to new technologies and enjoy innovation.
The aim is to significantly improve the quality of diagnoses using artificial intelligence. And to assist an overloaded system where radiologists are not available, for example during night shifts or in small hospitals. It also plays a key role when it comes to prevention. "Any finding that is not detected early enough can subsequently develop into something much bigger. A sick patient then costs the healthcare system more money," recalls Daniel Kvak.
One of the key things that sets Carebot apart from the competition is its ease of implementation. "The fact that we are able to plug anywhere, into any radiologist's or doctor's computer, into any system, in a very short time. Telemedicine is on the rise, we have also created a web application that doctors can use in the same way," says Daniel Kvak.
Read the full article here: https://www.deloitte.cz/fast50/?story=carebot