COVID-19 Detection by Applying Deep Learning on CT Images


Early detection and diagnosis are crucial factors to control the spread of the COVID-19 virus. Various deep learning-based methodologies have been proposed recently for COVID-19 screening in CT scans as a tool that can automate the diagnosis of the COVID-19 virus.

CT Images Helps with the Diagnosis

As compared to RT-PCR, the Thorax Computer Tomography (CT) is perhaps more dependable, helpful, and faster technology for the assessment and classification of COVID-19. Most hospitals have CT image screening; hence, the thorax CT pictures can be utilized for the early identification and detection of COVID-19 patients.

However, the COVID-19 classification based on the thorax CT mandates a radiology expert, and a lot of valuable time goes in screening the images. Presently, the COVID-19 test results take more than 24 hours to detect the virus in the human body. There is an immediate need to identify the illness in the early stage and immediately quarantine the infected individuals as no specific drugs are available to treat COVID-19.

How Deep Learning can Help?

Deep learning is the most effective technique that can be used in the field of medical science. It is a quick and efficient method for the analysis and prediction of various illnesses with an excellent accuracy rate. There are explicitly trained models to categorize the inputs into various categories desired by the programmers. In the medical field, they are used to identify heart problems, tumors using image analysis, diagnosing cancer, and many other applications.

It is also used to distinguish the CT scan images of the patients infected with the COVID-19 virus as positive or negative i.e. not infected. A self-developed model CTnet-10 was created that had an accuracy rate of 82.1%. To boost the accuracy, the CT scan image was also passed through several pre-existing models. It has been found that the VGG-19 model is best to categorize the images as COVID-19 positive or negative as it provides a better accuracy of 94.52%.

A CT provides a clear and speedy window into this process of diagnosing the virus, and deep learning of large multinational CT data could offer automated and reproducible biomarkers for classification and quantification of COVID-19 disease.

Investigators from National Institutes of Health (NIH) and NVIDIA have tried to create and evaluate a deep learning algorithm to detect COVID-19 on chest CT by utilizing the data from a globally diverse, multi-institutional dataset. The team received the COVID-19 CT scans from four hospitals across Italy, China, and Japan, where there was a broad variety in clinical timing and practice for CT use in outbreak settings.

The first model used was a segmentation model that defined the lung regions which were subsequently used by the classification model. Initially, the team created two classification models – one using the entire lung region with fixed input size (full 3D), and one using an average score of multiple regions within each lung at fixed image resolution (hybrid 3D).

When distinguishing between COVID-19 and other conditions, the hybrid 3D model achieved validation accuracy of 92.4 percent, while the full 3D model achieved an accuracy of 91.7 percent.


As the models were able to distinguish between COVID-19 and other types of pneumonia there may be a role for AI as one element of a CT-enhanced diagnosis, researchers have concluded. Subsequent models could include resource allocation, point of care detection for isolation of asymptomatic patients, or monitoring for response in clinical trials for medical countermeasures.