Automated Microaneurysm Detection in Retinal Fundus Images Based on the Combination of Three Detectors

作者：Hatanaka Yuji^*; Inoue Tsuyoshi; Ogohara Kazunori; Okumura Susumu; Muramatsu Chisako; Fujita Hiroshi

来源：Journal of Medical Imaging and Health Informatics, 2018, 8(5): 1096-1102.

DOI：10.1166/jmihi.2018.2419

• Microaneurysm Detection
• Diabetic Retinopathy
• Computer-Aided Detection
• Retinal Fundus Image
• Combination Detectors

Diabetic retinopathy is a complication of diabetes and is the leading cause of visual loss. Visual loss can be prevented by the early detection and treatment of diabetic retinopathy, which can be effectively detected by the fundus examination. Because microaneurysms are an early sign of diabetic retinopathy, several research groups have been developing automated microaneurysm detection methods. However, their performances are not sufficiently good or they rely on a database. Therefore, this paper proposes a method based on three types of microaneurysm detectors with a high generalization capability. An evaluation of combination methods using the Retinopathy Online Challenge database determined that the best method was based on the combination of a double-ring filter and a Shape Index based on the Hessian matrix. By comparing several previous methods, the proposed method also showed good performance. Moreover, the proposed method also evaluated using a different public database, the Standard Diabetic Retinopathy Database, calibration level 1. The proposed method showed a superior performance, although the performances of several previous methods was lower than them of ROC database. The proposed method exhibited flexibility with difference databases because this combination method was based on a simple majority rule.

• 出版日期2018-6