The Coincidental Findings of Lung Nodules with AI-assisted Chest X-Rays in the Tuberculosis Screening Project at Phra Nakhon Si Ayutthaya District

Abstract

Objective: Thailand implements a policy for pulmonary tuberculosis screening among patients with chronic diseases, but lung cancer screening is not included in the national policy. Since chest X-ray can reveal a variety of abnormalities. In addition to tuberculosis, lung nodules may also be found. This study determined the cancer detection rate and describe outcomes of an AI-assisted chest X-ray screening program integrated within community-based pulmonary tuberculosis screening. Methods: A retrospective descriptive study was conducted among individuals with chronic diseases who attended the pulmonary tuberculosis screening project in Phra Nakhon Si Ayutthaya District, which was implemented from February to May 2024. When chest X-ray images were obtained from the mobile X-ray unit-vehicle, the analysis process was performed: the first step involved a family physician using an AI-assisted Chest X-ray. If any abnormalities are found, the second step is to refer the results to a radiologist for official interpretation. In case of a suspected malignant lung nodule, the radiologist will recommend a CT scan or a follow-up. Results:  A total of 2,597 individuals with chronic diseases were screened for tuberculosis, comprising 816 males (31.4%) and 1,781 females (68.6%), with an average age of 62.87 years (male average, 62.11 years; female average, 63.22 years). Indeterminate Pulmonary Nodules (IPN) and granulomatous nodules were detected in various sizes and types, totaling 77 (2.96%). The radiologist recommended a chest CT scan for 39 (1.50%). After a 1-year follow-up, 23 (0.89%) underwent chest CT scans as recommended, and 4 participants were diagnosed with lung cancer, yielding a Cancer Detection Rate of 1.54 per 1,000. Conclusion: Lung cancer screening integrated with tuberculosis screening utilizing AI-assisted chest X-rays has the potential to be used for community screening. However, the steps taken after the first doctor uses AI to read the results and identify abnormalities in the lung nodules may be adjusted according to the appropriateness of the resources, which differ in each area and level of healthcare services. For example, regional hospitals and general hospitals may have a radiologist on staff to assist in interpreting the results. In community hospitals or primary care units, after a general practitioner or family physician reads the results using AI and identifies abnormalities in the case of lung nodules, a referral system should be in place to perform a low-dose CT scan. Analyzing chest X-rays with AI from the first step will increase confidence and reduce the likelihood of misdiagnosis for general practitioners and family physicians, who are the primary point of contact for cancer screening services. However, some patients may lose access to a low-dose CT scan due to personal or systemic limitations. In preparing the lung cancer screening system, the screening campaign period, equipment (including mobile X-ray units and vehicles, as well as artificial intelligence systems), referral systems, service units, and health personnel who will provide further care, both in cases of lung cancer diagnosis and lung nodule follow-up, must be prepared. A campaign should also be launched to raise public awareness of lung cancer and recognize the importance of receiving lung cancer screenings.