Artificial intelligence (AI) system created by Australian researchers can now quickly and accurately identify the COVID-19 virus from chest X-rays with over 98% accuracy. This process is more advantageous than the RT-PCR test that is presently in use.
Real-time polymerase chain reaction (RT-PCR), the most used COVID-19 test, can be expensive, time-consuming, and prone to false negative results. Radiologists must manually review CT images or X-rays to confirm a diagnosis, which can be error-prone and time-consuming, according to Professor Gandomi.
The new artificial intelligence method can be beneficial in nations with high COVID-19 infection rates and radiology shortages. Compared to CT scans, chest X-rays are more readily available, portable, and expose users to less ionizing radiation.
Fever, coughing, difficulty in breathing, and sore throat are typical signs of COVID-19, however, it can be challenging to differentiate it from the flu and other forms of pneumonia.
The new AI system can rapidly and reliably differentiate between COVID-19 instances, normal cases, and pneumonia in X-ray pictures. It uses a deep learning-based technique called a Custom Convolutional Neural Network (Custom-CNN).
“Deep learning provides a comprehensive approach, doing away with the necessity for manual biomarker discovery. According to Professor Gandomi, the Custom-CNN model expedites the identification procedure and offers a quicker and more precise diagnosis of COVID-19.
“If an RT-PCR test or rapid antigen test shows a negative result, due to low sensitivity, patients may require further examination via radiological imaging to confirm or rule out the virus’s presence. The new AI system may be useful in this case,” he continued.
Accuracy was the performance criteria used in a thorough comparison study to assess the Custom-CNN model. The researchers reported that the new model performs better than the existing AI diagnostic models based on the results.
