Rapid Radiology-Pathology Correlation using the TIES NLP System with Manual Abstraction Tools

Diagnostic imaging has provided clinicians with increasingly sophisticated noninvasive representations of disease. Correlation of radiological interpretations to histopathologic diagnoses has been an important element for quality improvement and education programs [1]. In the field of breast imaging, correlation to pathology findings is mandated by law [2]. Traditional manual methods for abstracting this information can be time consuming and expensive. While current electronic medical record systems increase the availability of coded data, the radiology and pathology data needed for correlation is typically stored as free-text. Consequently, many health systems and practices must manually correlate mammography to subsequent pathologic findings, which is time consuming, expensive, and typically produces a prolonged delay in assessment. We describe our experience in developing a method to rapidly correlate radiology-pathology data from Breast Imaging-Reporting and Data System (BIRADS) categories and pathology diagnoses. As more robust clinical information processing tools become available, these correlations can also become more automated, ultimately benefitting both disease surveillance and population health strategies.