As COVID-19 spreads exponentially, CT imaging departments are under significant pressures as it normally takes 5-15 minutes for a doctor to interpret each CT scan.
This technology for CT image analysis can assist in identifying characteristics of coronavirus pneumonia in CT scans with about 96% accuracy, and the entire test only takes 3 to 4 seconds. It is at least 60 times faster than human detection, therefore, making the virus detection procedure more efficient while maintaining high accuracy. More than 160 public institutions in China are currently using this technology. As of March 14, 2020, the system has already analyzed more than 240,000 CT image volumes (around 13,000 per day on average).
Technology for CT Image Analysis
This technology for CT image analysis, powered by deep learning algorithms, was trained on 5000 CT volumes and has been tested in hospitals throughout China. The process of CT Image Analysis training includes training and analyzing. It learns the differences between COVID-19 pneumonia, common pneumonia, and other situations. It is able to predict the probability of COVID-19 pneumonia and common pneumonia based on the input CT images. This CT Image Analysis technology can also output the lesion masks and affected lung volume ratio, helping doctors to effectively measure the development or treatment of COVID-19 patients.
Accuracy and Precision
Algorithm Accuracy Index
● Accuracy = 0.96
● F1-Score = 0.97
● Precision = 0.97
● Recall/Sensitivity = 0.98
● Specificity = 0.98
Algorithm Speed Index
● Thin-section chest CT scan in about 10 seconds.
● Thick-section chest CT scan in about 5 seconds.