News Photo

Solution that recognizes Covid-19 based on X-rays and CT scans

Coronavirus has revealed many weaknesses in today's health care systems. The latter perhaps has the highest priority in the case of global pandemic, so any decision in this area is "worth its weight in gold".

It’s been a while since the outbreak of coronavirus in China, and this has allowed creating necessary tests to determine if a person is sick or not. Despite the fact that morbidity peaks are behind us, health care systems need to draw conclusions and be prepared for the next possible waves.

In the situation of the growing number of infected and lack of tests the doctors often have to rely on other methods of diagnosis. One of the possible examples could be X-rays or CT scans. This practice is quite common when it comes to diagnosing pneumonia, abscesses, or enlargement of the lymph nodes. Given that X-ray machines are available in most hospitals, it is possible to carry out diagnostics without using tests.

Our company DataObrii, has developed a model that can recognize chest X-rays or CT scans and by using a computer vision to indicate a coronavirus. The model is based on Convolutional Neural Network Approach.

At that moment, the model recognizes pneumonia with 95% accuracy, and with 80% accuracy determines whether the inflammation is caused by COVD-19, pneumocystis, streptococcus, or SARS. Based on developed technology, we believe that our solution will help doctors prioritize screening for the most affected patients.

If the model determines that the X-ray image shows clean lungs, the patient isn’t infected and doesn’t have to be tested further with 99% accuracy. Otherwise, doctors will have to examine the patient and apply the necessary quarantine measures.

It’s worth noting that we add the block for the current and already existing patient’s symptoms. This will help doctors see full information and make a more balanced decision.

To make the solution accessible, portable, and easy in use, we have partnered with Memento Database - a powerful tool for storing and working with any kind of data. The integrated solution works on the android platform and has a nice interface and understandable navigation in desktop and mobile versions. Given that, hospitals and clinics have all the necessary equipment for fast and easy detection of COVID-19.

We should note that our solution is completely safe and secure. The system linked with the private cloud, so the data could be stored either there or on your facilities.

Let’s see how the app looks like.

The first steps are trivial for each application - download MementoDB and create an account. Then you will see the CDetector (name of the solution) - go there and fill in all possible fields (Figure 1).

Figure 1. CDetector app (desktop version)

When all the information about the patient's symptoms was entered and saved, the model will provide the result on the last tab -  ‘Auto diagnosis’:

  • General_pneumonia - indicates that the lungs are normal (Normal) or have a suspected illness (Pneumonia);
  • Cause of pneumonia - states the cause of pneumonia;

Prob means probability (1 = 100%):

  • Prob_normal - the probability that a person is healthy.

  • Prob_covid - the probability that a person is sick with a coronavirus

  • Prob_pneumocystis - the likelihood of pneumocystis

  • Prob_steptoccocus - streptococcus probability.

  • Prob_sars - the probability of SARS (Severe Acute Respiratory Syndrome).

In addition, we should state the following benefits of the CDetector:

1. The solution is a distributed system that allows you to build a Patient’s Database at the hospital. It is possible to share the access with many doctors, which helps organize the work and build a chain: one person makes an initial analysis, then another does an X-ray, then a third prescribes treatment. Such systematicity and orderliness can be implemented much faster into the existing structure of the hospital's functioning.

2. The system is very flexible. Since it is synchronized with Google Drive, it allows you to quickly change the configuration of the system to the individual needs of the hospital.

Besides that, we are currently working on a similar solution, which could assist doctors with diagnostics of respiratory disease. The solution will analyze a patient's breath via the audio channel and display the name of the most fitted disease.

To conclude, our big aim is to live in an easy, safe, and digital world. And in such challenging times, we decided to facilitate the work of the healthcare system through a data-driven solution. So, if your clinic or hospital needs this solution and you want to try it, please follow this link and fill in the form! Our team will assist you on each step.

Want to try? Fill in the form!

Share This News


Ready to find out more?