Automatic diagnosis of pathological myopia from heterogeneous biomedical data

• Main Authors: Zhang, Zhuo, Xu, Yanwu, Liu, Jiang, Wong, Damon Wing Kee, Kwoh, Chee Keong, Saw, Seang-Mei, Wong, Tien Yin
• Other Authors: School of Computer Engineering
• Format: Article
• Language: English
• Published: 2013
• Subjects: DRNTU::Engineering::Computer science and engineering
• Online Access: https://hdl.handle.net/10356/96436; http://hdl.handle.net/10220/11925
• Institution: Nanyang Technological University

Pathological myopia is one of the leading causes of blindness worldwide. The condition is particularly prevalent in Asia. Unlike myopia, pathological myopia is accompanied by degenerative changes in the retina, which if left untreated can lead to irrecoverable vision loss. The accurate diagnosis of pathological myopia will enable timely intervention and facilitate better disease management to slow down the progression of the disease. Current methods of assessment typically consider only one type of data, such as that from retinal imaging. However, different kinds of data, including that of genetic, demographic and clinical information, may contain different and independent information, which can provide different perspectives on the visually observable, genetic or environmental mechanisms for the disease. The combination of these potentially complementary pieces of information can enhance the understanding of the disease, providing a holistic appreciation of the multiple risks factors as well as improving the detection outcomes. In this study, we propose a computer-aided diagnosis framework for Pathological Myopia diagnosis through Biomedical and Image Informatics(PM-BMII). Through the use of multiple kernel learning (MKL) methods, PM-BMII intelligently fuses heterogeneous biomedical information to improve the accuracy of disease diagnosis. Data from 2,258 subjects of a population-based study, in which demographic and clinical information, retinal fundus imaging data and genotyping data were collected, are used to evaluate the proposed framework. The experimental results show that PM-BMII achieves an AUC of 0.888, outperforming the detection results from the use of demographic and clinical information 0.607 (increase 46.3%,p<0:005), genotyping data 0.774 (increase 14.7%, p<0.005) or imaging data 0.852 (increase 4.2%, p=0.19) alone. The accuracy of the results obtained demonstrates the feasibility of using heterogeneous data for improved disease diagnosis through our proposed PM-BMII framework.