A Proposal of Measuring Method for the Size of Abnormal Part Using Correlation Coefficient on Intensity Distribution from Optical Coherence Tomography Images

In ophthalmology, optical coherence tomography (OCT) is rapidly becoming popular to diagnose retinal diseases. However, the quantitative evaluation of retinal diseases has not been established for OCT image. Some clinical doctors desire a computer-aided diagnosis system to evaluate the size and the degree of disease quantitatively. In previous researches, the size of the retinal diseases is measured as abnormal parts of small regions in the number of layer boundaries, which is estimated by analysis method on a gray level profile for an OCT image. However, in an abnormal part, the difference of the number of layer boundaries is small between abnormal parts and normal parts, and the previous method can’t detect the size of the abnormal part. In this research, we propose a new method to measure the size of abnormal parts using correlation coefficient in gray level distribution for an OCT image. In the gray level distributions of the OCT image, the correlation coefficient is large in normal parts, and the coefficient is small between normal regions and abnormal parts. Therefore, our system can detect the position of borders between abnormal parts and normal parts using the distribution of correlation coefficient. We apply the proposed method to some retinal disease images. For an OCT image, we obtained the overlap rate 92.3[%] between the identified part by the proposed method and the clinical doctor for the abnormal part. The overlap rate is improved for 21 OCT images among 25 OCT images, and the improved rate in the proposed method is 84[%] compared with the previous metho