AI's ability to predict lung cancer in non-smokers

According to American cardiologist Eric Topol, AI in medicine aims to provide comprehensive views of medical data, improve decision-making, avoid errors, and assist in test interpretation and treatment recommendations.

Lung cancer caused 1.8 million deaths in 2020, with projections of over 28 million cases in 2040. The five-year survival rate for lung cancer in Europe is only 13%, with 20% of cases diagnosed at stage-I.

Early detection is crucial, as survival rates vary greatly depending on the stage of the disease. While various screening methods have been tried, low-dose computed tomography (LDCT) has shown the most significant reduction in lung cancer-related mortality.

AI can play a role in improving lung cancer identification and treatment. Combining AI systems with clinical and biomedical data can enhance screening and interpretation of lung images and other biomarkers, leading to better diagnosis.

Researchers at MIT and Massachusetts General Cancer Centre have developed an AI oracle that predicts a person's likelihood of developing lung cancer within the next six years. This approach addresses the limitations of existing guidelines that focus on smokers. The AI tool aims to increase screening rates and improve early detection.

The AI achieved an average AUC of 0.91 in identifying cancer within a year, with the highest score of 0.94 in the mixed data set from Taiwan. Its predictions for the next six years had an average AUC of 0.79, indicating promising results.

In conclusion, AI has the potential to assist in predicting lung cancer risk and improving screening rates, leading to better outcomes in lung cancer diagnosis and treatment.