The VA Connecticut Healthcare System implemented a cancer care tracking system that is used to track potentially malignant radiographic abnormalities. Vijay Garla, PhD, a postdoctoral researcher at Yale University, will describe how he and colleagues developed a natural language processing system to automatically code cancer alerts. The system extracts data, applies relevant decision rules that represent clinical practice guidelines, and flags reports for tracking that recommend surveillance of suspicious lesions.
The software performed well when tested with 60 days of chest CT reports, achieving sensitivity of 97% for lung node detection in 421 reports and specificity of 83%. An analysis of liver mass detection accuracy using 100 abdominal CT, MRI, and ultrasound reports produced sensitivity of 84% and specificity of 92%.
The tracking tool was deployed in 2011. The VA Connecticut Healthcare System is working with other VA medical centers to implement the technology.