JMIR Publications

ABSTRACT

Background:

Artificial intelligence can integrate complex features and may be used to predict the risk of developing lung cancer, thereby decreasing the need for unnecessary and expensive diagnostic interventions.

Objective:

Using electronic medical records to pre-screening patient’s risk for developing lung cancer.

Methods:

Two million participants were randomly selected from the Taiwan National Health Insurance Research Database from 1999 to 2013; We built a predictive lung cancer screening model with neural networks that were trained and validated using pre-2012 data and tested prospectively on post-2012 data. An age- and gender-matched subgroup that is 10 times larger than the original lung cancer group was used to assess the predictive power of EMR. Discrimination (area under the curve [AUC]) and calibration analyses were performed.

Results:

The analysis included 11,617 cases of lung cancer and 1,423,154 controls. The model achieved an AUC of 0.90 for the overall population and 0.87 in patients >55 years of age. The AUC in the matched subgroup was 0.82. The positive predictive value was highest (14.3%) among those >55-years-old with a preexisting history of lung disease.

Conclusions:

Our model achieved excellent performance at predicting lung cancer within one year and may be deployed for digital patient screening. Deep learning facilitates the effective use of EMRs to identify individuals at high risk for developing lung cancer.