When COVID-19 became a global pandemic, the virus’s rapid spread put immense pressure on healthcare systems to address the shortage of testing kits and delays in diagnosis. In such a critical situation, artificial intelligence (AI) has emerged as a powerful tool that can help detect and diagnose COVID-19.
The most widely used approach for detecting COVID-19 is real-time polymerase chain reaction (PCR). However, it has several limitations, such as high cost, lengthy turnaround time for results, and the likelihood of false-negative results because of limited sensitivity.
To address these concerns, health experts use additional technologies like computed tomography (CT) or X-rays for diagnosing the disease. Chest X-rays are more commonly used than CT scans because of the widespread availability of X-ray machines, lower cost of equipment, and lower exposure to ionizing radiation.
COVID-19 infection presents specific radiological biomarkers that can be observed through chest X-rays, so radiologists must manually search for these biomarkers. Such a process is time-consuming and prone to errors, which can pose challenges since the common symptoms of COVID-19 make it difficult to distinguish from flu and other types of pneumonia. Because of this, there is a need to develop an automated system for evaluating chest X-rays.
At the University of Technology, Sydney, researchers have developed an AI system that can quickly detect COVID-19 infection from chest X-rays with over 98% accuracy. The results of their study are discussed in the paper Auto-detection of the coronavirus disease by using deep convolutional neural networks and X-ray photographs.
The new system uses Custom Convolutional Neural Network (Custom-CNN), a deep learning-based algorithm that can quickly and accurately distinguish between COVID-19 cases and pneumonia in X-ray images. The performance of this model was evaluated through a comprehensive comparative analysis, using accuracy as the performance criterion. The results revealed that the new Custom-CNN model outperforms the other AI diagnostic models.
According to corresponding author Professor Amir H. Gandomi, deep learning provides an end-to-end solution by eliminating the need to search for biomarkers manually. Since it streamlines the detection process, the Custom-CNN models offer a faster and more accurate diagnosis of COVID-19.
Suppose a negative or inconclusive result is shown in a PCR or rapid antigen test because of low sensitivity. In that case, patients may undergo further examination through radiological imaging, which will confirm or rule out the presence of SARS-CoV-2. In this situation, the new AI system could be very beneficial.
The researchers acknowledge the crucial role radiologists play in medical diagnosis, but they also believe that AI technology can assist them in making more accurate and efficient diagnoses. The new system can also be particularly beneficial in countries that experience high levels of COVID-19, where there is a shortage of radiologists.
Through fast and accurate diagnosis of COVID-19, patients can receive the correct treatment, which will work best if taken within five days of the onset of symptoms. It can also help them isolate and protect others from infection, reducing the pandemic outbreaks.
