Coronavirus disease-2019 (COVID-19) originated in the Wuhan, Hubei Province, China in November 2019 and has since been declared a pandemic by the WHO. COVID-19 is an acute infectious disease, primarily affecting the respiratory system. Currently, real-time reverse transcription polymerase chain reaction (RT-PCR) performed on respiratory specimens is considered the
In general, combined chest CT, clinical symptoms, and laboratory tests facilitates the diagnosis of COVID-19. Increasing in-depth understanding of the disease, research, and the continuous improvement of artificial intelligence technology will further promote the establishment of a comprehensive prevention and control system of early screening, diagnosis, isolation, and treatment of COVID-19 pneumonia.
Some researchers have tried to apply artificial intelligence in CT image analysis to differentiate COVID-19 from other viral pneumonia patients. With clinical symptoms, laboratory testing results, and contact or travel history, the artificial intelligence system can help doctors identify patients with risk of progressing to a more severe disease state at the time of admission, for timely, precise, and effective treatment decisions. Hence, precise lesion labelling, segmentation, and quantification analysis of COVID-19 lesions is the future of artificial intelligence.
The artificial intelegence – system has outstanding perfomance in the detection of subtle GGO, which is the most easily missed typical CT feature of COVID-19. Also, it can precisely segment the lesion region, calculate the lesion volume, volume rates of lesions to total/left/right lung, and each in lung lobe. Comparing CT scans of the same patients at several time points, the radiologist can use the system to measuire changes in each lesion and track the progression of the disease.
Application of artificial intelligence (AI) in COVID-19
A large amount of CT images makes it difficult for radiologists to compare among serial studies. Thus, rapid detection, accurate location of lesions, and evaluation of lesion size, properties, and lesion dynamics are urgent issues that need to be addressed. An AI-assisted diagnostic system for COVID-19 has been developed in China. It takes about 15 s with an accuracy rate above 90%.
Streptococcus pneumonia is characterized by the consolidation of lobes or lobules without GGO. Both mycoplasma and aspiration peumonia distribute along bronchovascular bundle, which is significantly different from COVID-19 pneumonia. In study comparing chest CT from 219 patients with COVID-19 pneumonia in China and 205 patients with other cause of viral pneumonia in the United States, COVID-19 pneumonia cases were more likely to have a peripheral distribution (80 versus 57 percent), and GGO (91 versus 68 percent). COVID-19 patients more frequently had multifocal involvement on CT, compared with unifocalinvolvement in SARS and MERS.
Differential diagnosis with other pneumonia
Although the imaging features of COVID-19 overlap with those of SARS and MERS, there are differences on imaging exams that set the COVID-19 pnemonia apart. It is essential to make a differential diagnosis for early identification of borderline patients and determination of the appropriate treatment. Viral pneumonia is characterized by alveolar wall edema and interstitial changes.
Other studies focusing on the evolution of COVID-19 pneumonia concluded that the lung segments showed marked changes with decreasing of GGOs and increasing of crazy-paving pattern and consolidation. Patients recovering from COVID-19 can be tracked with CT to assess for long-term or permanent lung damage, including fibrosis.
Pan et al. investigated the time course of lung changes during recovery from COVID-19 pneumonia. The results showed that with the evolution of the disease, GGO would enlarge with a crazy paving pattern and partial consolidation. Lung involvement reached a maximum at about the 10th day after the onset, and the crazy-paving pattern was resolved entirely in the absorption stage about the 14th day after the disease onset, which means that the crazy-paving pattern may become an important indicator for evaluation of disease evolution.
Dynamic changes on chest CT
Chest CT can also evaluate the time course of COVID-19 and assess the evolution of disease severity. Chung et al. evaluated each of the five lung lobes and determined the degree of involvement and rated the severity of each lobe. The patients who had the highest score were admitted to the intensive care unit with multiple bilateral GGOs and subsegmental consolidation.