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 worldwide^1,2,3,4,5. 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 real-time reverse transcriptase-polymerase chain reaction (RT-PCR) assay, which remains the gold standard for COVID-19 diagnosis^{7,8,9,10,11,12,13,14,15,16,17,18,19,20,21,22,23}.

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 devices^{19,24,25,26,27}. Additionally, imaging classifications and structured reporting were guided by the characteristic imaging appearances associated with COVID-19²⁸. Imaging played a crucial role in the management of COVID-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 patients^{24,25,26,27,29,30,31}. 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 preparations, planning and management can be readily made for hospitals and clinics. To that purposes, the use of commercially available artificial intelligence (AI) software could be of potential benefit, especially in situations where medical resources may be limited. In this work, we were utilizing a pre-trained, commercial AI software. Previously, the AI software was trained with ground-glass nodules, and opacities, the algorithm’s accuracy in detecting pulmonary nodules was validated through multiple studies, including a single-center and multi-reader observer performance evaluation and a global multi-center observer performance study, respectively^32,33. The AI software, however, was adapted to the diagnostic workflow of a tertiary hospital within the settings specifically designated by the government for COVID-19, e.g., hospital emergency rooms as well as screening clinics (triage tents outside main hospital), outpatient safe clinics or community treatment centers for rapid screening of suspected patients with imaging appearances of pneumonia to the institution^31,34. Although numerous studies have developed new AI CAD systems specifically trained on COVID-19 chest X-ray datasets, there has been a lack of research on applying pre-existing, commercially available AI CAD systems—originally developed for other purposes such as lung nodule detection—to COVID-19 screening in an emergency context. This study seeks to fill that gap by exploring whether such a system can be repurposed for pandemic response without retraining. Related studies^35,36,37,38 have demonstrated promising performance with AI models tailored for COVID-19, but the unique contribution of this work lies in evaluating the applicability of a ready-to-use AI solution under urgent, resource-constrained circumstances. This study aims to provide insights into the cooperative experiences of adapting commercialized AI software for COVID-19 detection, assessing its generalizability using public databases, and discussing its implications for future pandemic preparedness and response^39,40,41.