A systematic review and meta-analysis of 140 studies revealed that artificial intelligence (AI) models, particularly those integrating imaging data, may outperform traditional regression models in predicting lung cancer risk. The findings indicated that AI models demonstrated higher accuracy, especially when utilizing low-dose computed tomography (LDCT) scans and achieved the highest accuracy for 1-year lung cancer incidence predictions. However, both AI and traditional models were associated with a high risk of bias, with AI models having a higher risk. Further validation across diverse populations and standardization of development practices are crucial before widespread implementation in clinical settings.

Source: JACR