Automated computerized analysis of cardiomegaly in digital chest radiographs

Cardiac size is an important diagnostic parameter, and cardiomegaly is often first diagnosed by radiologists during interpretation of chest radiographs. The authors are developing an automated method to determine the heart size in chest radiographs, based on the detection of lateral heart boundaries. Using 400 posteroanterior chest radiographs including various degree of cardiomegaly, the authors computed parameters related to the heart and lung size including the cardiac transverse diameter, the area of the heart, and the cardiothoracic ratio. Relationships between computed parameters and cardiomegaly were studied by means of receiver operating characteristics and multivariate analysis. Results indicate that the use of two parameters rather than one improved the detection of cardiomegaly compared with a single parameter