- Researchers at Rensselaer Polytechnic Institute have designed an algorithm with the ability to predict the need for ICU intervention in COVID-19 pneumonia patients
- The AI algorithm was tested on 295 participants from across the U.S., Iran, and Italy.
- According to the results AI accurately predicted 96% of cases that required ICU admission
According to the World Health Organization (WHO), 1 in every 5 people with COVID-19 require hospital admission. COVID-19 pneumonia although rare, is a serious complication of the novel coronavirus infection and a common cause of intensive care unit (ICU) admission. Pneumonia results from inflammation and destructions of lung alveoli. This presents as chills, shortness of breath, chest pain, or fatigue.
Furthermore, COVID-19 pneumonia can swiftly progress to lung failure and death. Early intervention is therefore a key factor in reducing pneumonia-associated fatality rate. A team of researchers at Rensselaer Polytechnic Institute have designed an AI-based method of assessing disease severity and predicting ICU admission.
ICU Admission in COVID-19 Pneumonia Cases
Doctors generally use Chest CT (computed tomography) to diagnose COVID-19 pneumonia. Especially in suspected patients with false-negative PCR results. Although CT is highly accurate it is often time-consuming. Recently researchers have been applying Artificial Intelligence (AI) methods for quicker analysis of CT images. However, non-imaging methods of assessing disease severity have not received the same amount of attention.
As a practitioner of AI, I do believe in its power. It really enables us to analyze a large quantity of data and also extract the features that may not be that obvious to the human eye.
Pingkun Yan, assistant professor of biomedical engineering at Rensselaer Polytechnic Institute
Therefore, the team of researchers combined CT images with non-imaging data such as vital signs, demographics, and laboratory results. The AI-based algorithm then combined all these data points and predicted whether the patient requires ICU admission.
Machine vs Human
The AI based Algorithm was tested on 295 patients hospitalized due to COVID-19. The patients belonged from across U.S., Italy and Iran. The researchers then compared their results to what kind of treatment the patient actually ended up needing.
According to the results published online in Medical Image Analysis, the AI model accurately predicted 96% of the cases that required ICU admission. Furthermore, it detected COVID-19 pneumonia in 90% of the cases. The authors of the study believe this is the first time a combination of imaging and non-imaging data has been used to predict patient outcome.
The Future of AI
The team is currently working on integrating the model for assessing other diseases. As the second wave surges on, such tools can be highly useful in reducing the disease burden of COVID-19.
We actually are seeing that the impact could go well beyond COVID diseases. For example, patients with other lung diseases. Assessing their heart disease condition, together with their lung condition, could better predict their mortality risk so that we can help them to manage their condition.
Pingkun Yan, assistant professor of biomedical engineering at Rensselaer Polytechnic Institute
Reference:
Chao, Hanqing, et al. “Integrative Analysis for COVID-19 Patient Outcome Prediction.” Medical Image Analysis, vol. 67, 2021, p. 101844., doi:10.1016/j.media.2020.101844.
