Researchers have developed a system using automated tools that can rapidly scan for COVID-19 with 98% accuracy, meaning the delays and errors of PCR testing can be curtailed.
Working out whether a patient has COVID-19, flu or other types of pneumonia can be difficult as symptoms such as fever, cough, breathing problems and a sore throat are common in all cases.
But this new system uses a custom convolutional neural network (Custom-CNN) – a form of deep learning-based algorithm – that can scan X-ray images and distinguish with a high degree of accuracy cases of COVID-19.
The findings of the research from the University of Technology Sydney (UTS) and Middle East University were published in Nature Scientific Reports on Monday.
Professor Amir H. Gandomi from UTS’ Data Science Institute said PCR tests, which were the most common way to confirm whether someone had contracted COVID-19, were expensive, slow and sometimes delivered false positive results.
“To confirm a diagnosis, radiologists need to manually examine CT scans or X-rays, which can be time-consuming and prone to error,” Gandomi said.
“The new AI system could be particularly beneficial in countries experiencing high levels of COVID-19 where there is a shortage of radiologists.”
The corresponding author of the paper entitled ‘Auto-detection of the coronavirus disease by using deep convolutional neural networks and X-ray photographs’ added that using chest X-rays to determine COVID-19 came with the benefit of being a portable way to bring tests to local areas and subjected patients to less exposure to ionising radiation than CT scans.
“If a PCR test or rapid antigen test shows a negative or inconclusive result, due to low sensitivity, patients may require further examination via radiological imaging to confirm or rule out the virus’s presence. In this situation the new AI system could prove beneficial,” Gandomi said.
“While radiologists play a crucial role in medical diagnosis, AI technology can assist them in making accurate and efficient diagnoses,” he said.
Given the impact of the pandemic on public health and on the global economy, accurate and speedy COVID-19 readings are important. The benefits of the Custom-CNN model for patients are also significant, with earlier diagnosis meaning people can be treated with antivirals that are most effective within five days of symptom onset.
The researchers evaluated their system using a comprehensive comparative analysis to show the model outperformed other AI diagnostic models.
“Deep learning offers an end-to-end solution, eliminating the need to manually search for biomarkers,” Gandomi said.
“The Custom-CNN model streamlines the detection process, providing a faster and more accurate diagnosis of COVID-19.”
