AI System for Diagnosis, Quantitative Measurements, and Prognosis of COVID-19 Pneumonia
The COVID-19 has emerged as pandemic since December, 2019. By April 24th, 2020, more than 2,750,000 patients were confirmed as novel coronavirus pneumonia (NCP), with a total death toll of 190,000 in the world, according to the statistics released by Johns Hopkins University. COVID-19 patients infected by SARS-CoV-2 virus develop pneumonia and rapidly progress to respiratory failure. However, rapid diagnosis and identification of high-risk patients for early intervention are challenging. On April 25th, 2020, Sun Yat-Sen Memorial Hospital, Macau University of Science and Technology, and Tsinghua University etc. published an article “Clinically Applicable AI System for Accurate Diagnosis, Quantitative Measurements, and Prognosis of COVID-19 Pneumonia Using Computed Tomography” online in “Cell”, proving an AI system of diagnosing early NCP. The vice-dean of our hospital, Professor. Tianxin Lin is the co-correspondence author, and the director of radiology department, Professor. Jun Shen is the co-first author.
Many affected NCP patients progress rapidly into severe acute respiratory failure with a very poor prognosis and high mortality. There are over 60% of patients died once they progressed into the severe or critical illness stage. Therefore, identifying risk factors and parameters that can allow the creation of an accurate prognosis predictive model and hopefully lead to improved clinical outcomes are critical in the planning of early intervention and intense monitoring.
Chest computed tomography (CT) is an important tool in the diagnosis of pneumonia. CT scanning procedure has a faster turnaround time than a molecular diagnostic test performed in a standard laboratory, can provide more detailed information related to the pathology, and is better for the quantitative measurement of lesion size and the extent or severity of lung involvement, which may have prognostic implications. Therefore, an accurate CT-based artificial intelligence (AI) system may have the potential to assist in the early diagnosis for planning, monitoring and treatment, and establishing the reference for longitudinal follow ups.
Using a large CT database from 3,777 patients, they developed an AI system based on a lung lesion segmentation model and a diagnosis analysis model. First, they had demonstrated that the AI system can differentiate NCP from other common pneumonia and normal controls by using multi-center data validation. Its performance was overall superior to that of junior radiologists and comparable to mid-senior radiologists. Then, the AI system has potential in the evaluation of drug treatment efficacy in an objective quantitative way. The AI-generated quantitative measurements can be used to evaluate the effect of drug treatment on lesion size and volume changes between pre-treatment and post-treatment. Finally, the AI system is conducive to the prediction of progression to critical illness. They found that lung lesions were associated with the respiratory system function and other organ failures. A combination of lung lesions and clinical metadata can contribute significantly to the prognosis prediction.
Significantly, the AI system can assist radiologists and physicians in performing a quick diagnosis and quantitative evaluation of drug treatment effects especially when the health system is overloaded. AI system identified important clinical markers that correlated with the NCP lesion properties. Together with the clinical data, the AI system was able to provide accurate clinical prognosis that can aid clinicians to consider appropriate early clinical management and allocate resources appropriately. They have made this AI system available globally to assist the clinicians to combat COVID-19.
Hyperlink of the article: https://www.cell.com/cell/fulltext/S0092-8674(20)30551-1?_returnURL=https%3A%2F%2Flinkinghub.elsevier.com%2Fretrieve%2Fpii%2FS0092867420305511%3Fshowall%3Dtrue
