AI integrations with lung cancer screening: Considerations in developing AI in a public health setting

Abstract

Lung cancer screening implementation has led to expanded imaging of the chest in older, tobacco-exposed populations. Growing numbers of screening cases are also found to have CT-detectable emphysema or elevated levels of coronary calcium, indicating the presence of coronary artery disease. Early interventions based on these additional findings, especially with coronary calcium, are emerging and follow established protocols. Given the pace of diagnostic innovation and the potential public health impact, it is timely to review issues in developing useful chest CT screening infrastructure as chest CT screening will soon involve millions of participants worldwide. Lung cancer screening succeeds because it detects curable, early primary lung cancer by characterizing and measuring changes in non-calcified, lung nodules in the size-range from 3mm to 15 mm in diameter. Therefore, close attention to imaging methodology is essential to lung screening success and similar image quality issues are required for reliable quantitative characterization of early emphysema and coronary artery disease. Today’s emergence of advanced image analysis using artificial intelligence (AI) is disrupting many aspects of medical imaging including chest CT screening. Given these emerging technological and volume trends, a major concern is how to balance the diverse needs of parties committed to building AI tools for precise, reproducible, and economical chest CT screening, while addressing the public health needs of screening participants receiving this service. A new consortium, the Alliance for Global Implementation of Lung and Cardiac Early Disease Detection and Treatment (AGILE^DxRx) is committed to facilitate broad, equitable implementation of multi-disciplinary, high quality chest CT screening using advanced computational tools at accessible cost.