Genetic U-Net: Automatically Designed Deep Networks for Retinal Vessel Segmentation Using a Genetic Algorithm

Recently, many methods based on hand-designed convolutional neural networks (CNNs) have achieved promising results in automatic retinal vessel segmentation. However, these CNNs remain constrained in capturing retinal vessels in complex fundus images. To improve their segmentation performance, these CNNs tend to have many parameters, which may lead to overfitting and high computational complexity. Moreover, the manual design of competitive CNNs is time-consuming and requires extensive empirical knowledge. Herein, a novel automated design method, called Genetic U-Net, is proposed to generate a U-shaped CNN that can achieve better retinal vessel segmentation but with fewer architecture-based parameters, thereby addressing the above issues. First, we devised a condensed but flexible search space based on a U-shaped encoder-decoder. Then, we used an improved genetic algorithm to identify better-performing architectures in the search space and investigated the possibility of finding a superior network architecture with fewer parameters. The experimental results show that the architecture obtained using the proposed method offered a superior performance with less than 1% of the number of the original U-Net parameters in particular and with significantly fewer parameters than other state-of-the-art models. Furthermore, through in-depth investigation of the experimental results, several effective operations and patterns of networks to generate superior retinal vessel segmentations were identified. The codes of this work are available at https://github.com/96jhwei/Genetic-U-Net.

I. Introduction

The retinal vascular system can be observed non-invasively in vivo in humans [1]. In addition, changes in the vasculature are often associated with certain diseases, leading ophthalmologists and other physicians to consider fundus examination a routine clinical examination [2], [3]. Many diseases can be diagnosed and tracked [2] by observing the retinal vascular system. Pathological changes in retinal vessels can reflect either ophthalmology diseases or other systemic diseases, such as wet age-related macular degeneration and diabetes [4]. Diabetic retinopathy can lead to the growth of new blood vessels, and atherosclerosis [5] associated with wet age-related macular degeneration can cause the narrowing of blood vessels. Moreover, the retinal vascular system of each eye is unique. Without pathological changes, it does not alter throughout the lifetime. Hence, observation of the retinal vascular system can also be applied in biometrics [6], [7]. Through retinal vessel segmentation, relevant morphological information of retinal vascular trees (such as the width, length, and curvature of blood vessels) can be obtained [8]. Consequently, precise retinal vessel segmentation is significant. However, owing to the complexities of retinal vascular structures, manual inspection is subjective, time-consuming, and laborious [9], [10]. Therefore, developing an effective algorithm for the automated segmentation of retinal vessels to support ophthalmologists in clinical assessment has been of great interest.