项目名称: 基于视觉注意模型的眼底病变自动识别方法研究
项目编号: No.61472102
项目类型: 面上项目
立项/批准年度: 2015
项目学科: 计算机科学学科
项目作者: 卜巍
作者单位: 哈尔滨工业大学
项目金额: 83万元
中文摘要: 我国是全世界盲人最多的国家。糖尿病视网膜病变、老年黄斑变性等眼底病是导致不可逆转性失明的主要原因。眼底病普查是及时诊治眼底病的有效手段,可有效减少失明的发生。目前,眼底病主要是通过专家人工检查眼底是否出现病变来诊断,这使得大范围的眼底病普查不能开展,大量眼底病患者因得不到及时诊断和治疗,导致视力受损甚至失明。为此,本课题将通过分析彩色眼底图像中各种病变的特点,把眼底病变作为显著性目标,研究基于视觉注意模型的眼底病变自动识别方法,实现对眼底病变的自动检测和识别。研究内容主要包括:(1)扩充已有的眼底病专业数据库,为眼底病变自动识别研究提供数据基础;(2)研究各种眼底病变的显著性特征提取与学习算法,建立适合眼底病变检测的视觉显著性计算模型;(3)研究各种眼底病变的分割算法;(4)研究能够准确区分多种眼底病变的识别算法。本项目旨在解决眼底病变人工检查的不足,提高眼底病变自动识别的效率和准确率。
中文关键词: 医学图像处理;计算机辅助诊断;视觉注意模型;特征提取;图像识别
英文摘要: China has the world's largest blind people. The fundus diseases, such as Diabetic Retinopathy, Age-related Macular Degeneration are the main cause of the blindness. At the present stage, the screening of fundus diseases is conducted by ophthalmologist manually, which is time-consuming, repetitive and tiring, and consequently errors can be made. Many patients has lost vision or even blind without the timely screening and treatment. This programme of research aims to develop automated methods for fundus lesions identification by using visual attention model. The main research contents include: (1) build a database with large-scale fundus images and data management system,(2) design novel approaches for lesion feature extraction and learning, develop the special visual saliency computing model to detect lesion regions on the fundus image, (3) propose the new segmentation algorithms to segment lesion regions, (4) propose the new classification algorithms for the lesion recognition. The research will improve the efficiency and precision of fundus screening.
英文关键词: Medical image processing;Computer aided diagnosis;Visual attention model;Feature extraction;Image recognition