Applying artificial intelligence to ensure high quality and equitable lung cancer screening.
Authors
Affiliations (5)
Affiliations (5)
- Department of Radiology, University of Iowa, Iowa City, IA, USA.
- Roy J. Carver Department of Biomedical Engineering, University of Iowa, Iowa City, IA, USA.
- Holden Comprehensive Cancer Center, University of Iowa, Iowa City, IA, USA.
- Department of Medicine, University of Pennsylvania Perelman School of Medicine, Philadelphia, Pennsylvania, PA, USA.
- Department of Internal Medicine, University of Iowa, Iowa City, IA, USA.
Abstract
Lung cancer screening (LCS) with low-dose computed tomography has the potential to improve early detection and promote more equitable health outcomes. However, traditional eligibility criteria, based primarily on age and smoking history, may overlook high-risk individuals, particularly in underrepresented populations. These include racial minorities and individuals living in rural areas, who often face limited access to screening centers and high-quality imaging interpretations. Artificial intelligence (AI) offers promising solutions to potentially enhance the effectiveness and equity of LCS. First, AI could refine risk stratification by incorporating additional clinical data, social determinants of health, environmental exposures, and comorbidities, thereby identifying high-risk individuals who may be missed by conventional criteria (e.g., Black Americans, women). Second, AI could improve access to high-quality screening by enhancing image acquisition across diverse technologies and enabling remote interpretation through telehealth. Third, AI tools could support radiologists by increasing the accuracy of nodule detection and improving the assessment of malignancy risk in detected nodules. Finally, AI could assist in managing incidental findings and facilitate opportunistic screening, further expanding the impact of LCS. Despite its promise, the implementation of AI in clinical practice faces several barriers. These include regulatory hurdles, the need for clinical billing codes, and substantial investment in infrastructure, training and ongoing monitoring of these technologies. Further, the consideration of fairness-aware frameworks to mitigate racial bias in AI tools developed from non-representative datasets. Integrating AI into the radiologic workflow, with attention to these challenges, may address disparities and improve the overall quality and reach of LCS. However, AI implementation will need to be carefully evaluated to determine whether it is achieving these goals.