HRF fellowship in artificial intelligence health research – health systems sustainability

Research overview

Lung cancer is the leading cause of cancer-related death worldwide. Lung cancer screening with CT imaging allows for early detection of lung cancer and has been shown to reduce deaths related to lung cancer by 20-24%. Lung cancer appears on CT as a lung nodule, and with current methods in clinical practice, it is not possible to determine with certainty whether a lung nodule represents lung cancer or whether it is benign.

Artificial intelligence (AI) holds the potential to analyze images and extract information beyond human perception. This study investigated AI to extract additional information from CT images to improve the classification of lung nodules as benign or malignant.

Based on the demonstrated ability of the AI algorithm to make accurate predictions regarding lung cancer, we developed a hybrid strategy to incorporate AI into patient care. The study found that the hybrid strategy resulted in earlier diagnoses of lung cancer. In addition, the strategy reduced the number of unnecessary follow-up investigations for lung nodule management, and substantial cost savings could be achieved by applying AI in lung cancer screening.

Research supported through the fellowship has contributed to an evidence base which will help further guide the development and implementation of AI to enable earlier lung cancer diagnosis, improved patient outcomes, and reduced healthcare costs. The fellowship helps increase capacity in the Canadian health research community and health system to advance AI to improve healthcare in Canada and around the world.

Dr. Scott Adams

Real-world applications

Current approaches to lung nodule management result in delayed diagnoses of lung cancer and costly unnecessary follow-up investigations such as CT, PET/CT, and biopsy. As lung cancer screening programs are developed and scaled across Canada and across the world—resulting in the potential for thousands of additional lung cancers to be detected at earlier stages—the challenge of ensuring accurate and timely lung cancer diagnosis becomes even more critical. As our results suggest, the integration of AI into lung cancer screening programs would allow physicians to diagnose lung cancer earlier, while minimizing the number of diagnostic tests performed for each patient and achieving substantial healthcare cost savings. Increased cost-effectiveness of lung cancer screening enabled by AI may allow screening programs to be scaled to a broader population, further reducing deaths related to lung cancer.