AI-Powered COVID-19 CT Algorithm Provides Quantitative Measurement for Suspected Coronavirus Patients

The latest updated version of an artificial intelligence (AI)-driven COVID-19 medical imaging solution can now help radiologists distinguish between coronavirus and other abnormalities, such as common pneumonia on chest CT scans.

RADLogics (New York, NY, USA) has unveiled the latest version of the company’s AI-Powered COVID-19 CT algorithm. Building on the company’s AI-Powered solution that has processed and analyzed hundreds of thousands of suspected coronavirus cases globally, the latest update delivers a complex deep learning system consisting of several models utilized to detect, localize and segment regions in the lungs infected with COVID-19. The AI-Powered solutions are poised to not only alleviate the increased burden associated with COVID-19, but to help support improved outcomes by reducing burnout and errors.

As part of RADLogics’ latest version of its algorithm, three different analyses can now be performed simultaneously on raw chest CT image scans including: 1) a lungs region-of-interest are cropped with lung abnormalities detected; 2) the lung lobes are segmented and; 3) if the nodules plug-in is activated, focal Ground Glass Opacities (GGOs) are detected. According to leading physicians, these measurements are key features in determining patient classification into COVID-19 and non-COVID-19 indications. The overall system produces the decision whether the case is suspected for COVID-19 with a confidence level (in percentages). These measurements along with other features are used by radiologists to distinguish between COVID-19 and other abnormalities such as common pneumonia.

To further validate the ability of AI to distinguish COVID-19 from other respiratory diseases, members of the RADLogics’ algorithm development team, led by Professor Hayit Greenspan from Tel Aviv University, studied a fully automated AI-based system that takes as input chest CT scans and triages COVID-19 cases. The study explored multiple descriptive features, including lung and infections statistics, texture, shape and location, to train a machine learning-based classifier that distinguishes between COVID-19 and other lung abnormalities (including community acquired pneumonia). The research evaluated the system on a dataset of 2,191 CT cases and demonstrated a robust outcome with 90.8% sensitivity at 85.4% specificity with 94.0% ROC-AUC.

“With the US in the midst of an unprecedented rise in COVID-19 infections, with current hospitalizations at an all-time record of more than 90,000 patients, there is an increasing need for AI solutions in medical imaging,” said Moshe Becker, CEO and Co-Founder of RADLogics. “Coronavirus-related infection rates are experiencing a sharp increase in most states - from rural communities to urban areas - that have the potential to overwhelm ER, ICU and radiology teams with a surge of patients, and AI-Powered medical imaging analysis solutions are poised to reduce this pressure through improved patient triage, monitoring and management.”

Related Links:
RADLogics