Cardio-Thoracic Ratio Measurement Using Non-linear Least Square Approximation and Local Minimum.

This paper presents a method to automatically measure cardio-thoracic ratio (CTR) from a chest radiographic images using non-linear least square approximation and local minimum. The proposed method consists of initial boundary point identification, cardiac diameter measurement, thoracic diameter measurement and cardio-thoracic ratio measurement. First, the initial boundary points used to approximate the region of thoracic cavity are identified using general human anatomy features. Then the non-linear least square approximation and local minimum are used to detect the heart boundary. Finally, the thoracic cage boundary is detected and the cardio-thoracic ratio can be measured. The proposed method is tested on a set of 255 chest radiographs. The experimental results are evaluated using correlation test between two sets of numerical measurement which are measured by our proposed method and by the radiologists. The evaluation reveals that the correlation result on CTR is about 78%.