Feature-transfer network and local background suppression for microaneurysm detection

  • PDF / 2,063,642 Bytes
  • 13 Pages / 595.276 x 790.866 pts Page_size
  • 62 Downloads / 178 Views

DOWNLOAD

REPORT


ORIGINAL PAPER

Feature-transfer network and local background suppression for microaneurysm detection Xinpeng Zhang1

· Jigang Wu1 · Min Meng1 · Yifei Sun1 · Weijun Sun2

Received: 20 July 2019 / Revised: 15 June 2020 / Accepted: 24 August 2020 © Springer-Verlag GmbH Germany, part of Springer Nature 2020

Abstract Microaneurysm (MA) is the earliest lesion of diabetic retinopathy (DR). Accurate detection of MA is helpful for the early diagnosis of DR. In this paper, an efficient approach is proposed to detect MA, based on feature-transfer network and local background suppression. In order to reduce noise, a feature-distance-based algorithm is proposed to suppress local background. The similarity matrix of feature distances is calculated to measure the difference between background noise and retinal objects. Moreover, a feature-transfer network is proposed to detect MAs with imbalanced data. For each training process, the optimized weights and bias are transferred to the next training, until the optimal network is generated. Experimental results demonstrate that the proposed approach can accurately detect subtle MAs surrounded by complex background. Furthermore, the sensitivity values on the public datasets are up to 98.3%, 100%, 99.3%, 100%, 96.5%, respectively. The proposed approach outperforms the state-of-the-arts, in terms of the competition performance measure score. Keywords Feature-transfer network · Local background suppression · Feature distance · Microaneurysm detection

1 Introduction Diabetes is a kind of chronic non-infectious metabolic diseases characterized by hyperglycemia. It has become a public health problem all over the world [1]. Diabetes causes many complications such as blindness, renal and cardiovascular diseases [2]. The number of diabetic patients in the world is estimated to exceed 700 million in 2045 [3]. Diabetic retinopathy (DR) is the most common micro-vascular complication of diabetes, resulting in the blindness among adults

B

Jigang Wu [email protected] Xinpeng Zhang [email protected] Min Meng [email protected] Yifei Sun [email protected] Weijun Sun [email protected]

1

School of Computer Science and Technology, Guangdong University of Technology (GDUT), Guangzhou, China

2

School of Automation, Guangdong University of Technology (GDUT), Guangzhou, China

aged 20 to 74 years [4]. About a third of peoples have signs of DR, and a third of these might have vision-threatening. The annual budget of governments for the diagnosis and treatment is enormous. The early precaution and screening can effectively reduce the rate of progression [5]. Microaneurysm (MA) is the earliest symptom of DR. MA generates a small swell on the vessel wall [6]. Artificial screening is inefficient and time-consuming. Therefore, image processing-based techniques are necessary for automatic MA detection [7]. However, some adverse factors decrease the precision of MA detection. Firstly, the contrast of subtle MAs is low. Morphology-based methods and filter-based methods are failure to recogniz