New app could be early diagnosis tool for COVID-19 in future

  • Date

    Thu 4 Jun 20

A black and white CT Scan.

A new app which assesses CT images of patients’ lungs could prove to be a useful tool in the future for frontline NHS staff dealing with suspected COVID-19 cases.

Developed by computer scientists at the University of Essex, the app is already showing a good rate of accuracy despite being at an early stage of development. 

The COVID-19 detection app works by a CT scan of a patient’s lungs being uploaded and then it will give the percentage of the risk of being infected by COVID-19. If successful, the app could prove a useful early diagnosis test for COVID-19 which would take under an hour from CT scan to result, compared to swab tests where results can take a minimum of 24 hours. 

The team from the School of Computer Science and Electronics Engineering working on the project are MSc student Suraj Ghuwalewala, under the guidance of Dr Haider Raza, an expert in artificial intelligence (AI) for decision making.  
The challenge for Suraj and Dr Raza is there is a lack of available CT scan data for COVID-19. They have used an open-source CT lung scan dataset approved by the senior radiologist in Tongji Hospital, China. However, they are hoping to partner with an NHS Trust to enrich the CT image dataset as including more images in training the AI model may improve the accuracy of the app, which works well in the theory of deep learning, which is a function of AI.

New app could be early diagnosis tool for COVID-19 in future

CT scans are already seen as a more reliable scan than x-rays to assess a COVID-19 patient as they offer more detailed information. If the app is eventually adopted by NHS Trusts, it could mean frontline staff would be able to quickly assess if a patient had COVID-19 if there was not a radiologist available. 
“Although this research is at a very early stage, we have seen some very impressive results on its accuracy,” explained Dr Raza. “If the virus reappears, our algorithm has great potential to be applied in clinical applications for accurate and rapid COVID-19 diagnosis, which would be a great help for frontline medical staff.”
Suraj added: “We are now looking to collaborate with medical experts and hospitals, so we can enrich our dataset of CT scan images, which will aid in developing the app and improving accuracy even further. With enough data, our algorithm can be expanded to diagnose other pneumonia-like diseases as well.”