Utilizing AI CAD for early pandemic screening in chest radiographs

Sep 30, 2026

Upload by Ophelia

Utilizing pre-trained commercial AI CAD software for diagnosing COVID-19 and other pneumonia-related chest radiograph findings, the study evaluates performance using a public dataset from the early pandemic phase. Reported sensitivity and specificity for pneumonia detection were 86.83% and 59.59%, respectively, with improved results using posterioranterior (PA) radiographs at 89.16% sensitivity and 67.41% specificity. Findings support AI CAD as an initial triage tool where advanced resources are limited, highlighting rapid adaptation of existing systems for outbreak response.

LanguageEnglish

CategoryHealthcare

FormatPDF

Pages9

Reading Time23 min

AccessFree to download

Introduction

COVID-19 diagnostic context and imaging role

Radiography, triage, and need for rapid screening

Methods

Repurposed pre-trained commercial AI CAD workflow

Results

Sensitivity and specificity for pneumonia and COVID-19-related findings

Impact of posterioranterior radiographs

[www. nature.com/scientificreports](www. nature.com/scientificreports) OPEN Utilizing AI CAD for early pandemic screening in chest radiographs Young Beom Kim To investigate the potential application of existing artificial intelligence (AI) software in diagnosing COVID-19 (coronavirus disease 2019) and other pneumonia-related radiographic findings with the unprecedented challenge by COVID-19 pandemic, leveraging pre-trained commercial AI software originally designed for detecting pulmonary nodules in chest radiographs, we assessed its performance using public dataset comprising chest radiographs collected during the early stages of pandemic. The software demonstrated promising results, achieving a sensitivity of 86.83% and a specificity of 59.59% in detecting pneumonia features, including those caused by COVID-19. Considering only posterioranterior (PA) radiographs, AI software exhibited improved performance, with sensitivity 89.16% and specificity 67.41%. The findings suggest the AI CAD software could serve as a triage for initial screening of infectious pulmonary diseases, especially in regions with limited access to advanced medical resources. This study underscores the importance of promptly adapting existing AI CAD to address urgent healthcare needs during infectious disease outbreaks. Keywords Artificial intelligence, Infectious pulmonary disease, Chest radiograph, Computer aided detection, Initial stage of pandemic The COVID-19 pandemic, caused by the novel coronavirus SARS-CoV-2, emerged in December 2019 in Wuhan, China. Recognizing the rapid spread and severity of the virus, the World Health Organization (WHO) swiftly declared the outbreak a Public Health Emergency of International Concern (PHEIC) in January 2020. By March 2020, the situation had escalated, leading the WHO to categorize COVID-19 as a pandemic, signaling an urgent need for global action and cooperation to mitigate the virus’s spread and impact on public health systems worldwide1–5. This declaration highlighted the necessity for immediate, coordinated responses across nations to control the pandemic and minimize its adverse effects. In response to the pandemic, the American College of Radiology (ACR) recommended viral nucleic acid testing as the primary diagnostic method for confirming COVID-19 cases. At the same time, they advised a cautious and judicious use of imaging modalities for clinical decision-making, particularly in regions with limited medical resources6. Various radiological findings related to COVID-19 have been reported using computed tomography (CT) and chest radiography (CXR) examinations. These findings have been correlated with molecular diagnostic tests, specifically the realtime reverse transcriptase-polymerase chain reaction (RT-PCR) assay, which remains the gold standard for COVID-19 diagnosis7–23. Guidelines for using imaging equipment were developed based on clinical consensus and the availability of local medical resources, including clinical staff, viral testing capabilities, personal protective equipment (PPE), hospital beds, ventilators, and portable imaging devices19,24–27. Additionally, imaging classificationsand structured reporting were guided by the characteristic imaging appearances associated with COVID-1928. Imaging played a crucial role in the management ofCOVID-19, especially in scenarios requiring rapid assessment of patients’ conditions, such as emergency rooms, screening clinics, outpatient safe clinics, and community treatment centers. Chest radiographs were particularly valuable for screening, follow-up, and establishing baseline conditions, assessing the severity and progression of the infection, identifying superimposed treatable processes, and distinguishing between COVID-19 and non-COVID-19 patients24–27,29–31. With the rapid surge of imaging volume during the pandemic, rapid screening, triaging, and isolation of COVID-19-positive or suspected patients became critical such that precautionary p

FAQs

What is the study’s main goal regarding AI CAD in COVID-19 screening?

To investigate whether existing, pre-trained commercial AI software can be repurposed to diagnose COVID-19 and other pneumonia-related findings on chest radiographs during the early pandemic stage without retraining.

How was the AI CAD performance evaluated?

The software was tested on a public dataset of chest radiographs collected in the early stages of the pandemic, measuring sensitivity and specificity for pneumonia detection, including COVID-19-caused features.

What performance did the AI CAD system achieve and how did PA radiographs affect it?

For pneumonia detection overall, sensitivity was 86.83% and specificity was 59.59%. Using posterioranterior (PA) radiographs improved performance to 89.16% sensitivity and 67.41% specificity.

Sep 30, 2026

Upload by Ophelia