Unsupervised Approaches for the Segmentation of Dry ARMD Lesions in Eye Fundus cSLO Images

Abstract

Age-related macular degeneration (ARMD), a major cause of sight impairment for elderly people, is still not well understood despite intensive research. Measuring the size of the lesions in the fundus is the main biomarker of the severity of the disease and as such is widely used in clinical trials yet only relies on manual segmentation. Artificial intelligence, in particular automatic image analysis based on neural networks, has a major role to play in better understanding the disease, by analyzing the intrinsic optical properties of dry ARMD lesions from patient images. In this paper, we propose a comparison of automatic segmentation methods (classical computer vision method, machine learning method and deep learning method) in an unsupervised context applied on cSLO IR images. Among the methods compared, we propose an adaptation of a fully convolutional network, called W-net, as an efficient method for the segmentation of ARMD lesions. Unlike supervised segmentation methods, our algorithm does not require annotated data which are very difficult to obtain in this application. Our method was tested on a dataset of 328 images and has shown to reach higher quality results than other compared unsupervised methods with a F1 score of 0.87, while having a more stable model, even though in some specific cases, texture/edges-based methods can produce relevant results.

Encoder-Weighted W-Net for Unsupervised Segmentation of Cervix Region in Colposcopy Images

“…During training, soft N-cut loss and reconstruction error are jointly minimized, and the encoded image is then post-processed to generate the final segmentation result. One study [ 29 ] adopted W-Net for the segmentation of confocal scanning laser ophthalmoscopy (cSLO) images. The main difference between the W-Net proposed in [ 29 ] and the baseline W-Net is that a pooling layer is added before calculating the soft N-cut loss, to reduce memory consumption.…”

“…One study [ 29 ] adopted W-Net for the segmentation of confocal scanning laser ophthalmoscopy (cSLO) images. The main difference between the W-Net proposed in [ 29 ] and the baseline W-Net is that a pooling layer is added before calculating the soft N-cut loss, to reduce memory consumption.…”

“…The larger the image, the higher the computational overhead. As a baseline model of our work, we adopt the W-Net architecture with an additional pooling layer before soft N-cut loss calculation [ 29 ]. The pooling layer reduces the output map size and therefore can reduce memory requirement.…”

Context-aware Attention U-Net for the segmentation of pores in Lamina Cribrosa using partial points annotation

“…Comparison with other state-of-the-art methods. We compare the proposed method with traditional active contour approach proposed by Chan Vese [22], the unsupervised method W-Net [23], as well as the supervised methods of original U-Net [9] and Attention U-Net [19]. We manually delineated a peripheral mask to only retain the most contrasted area of the LC for Chan Vese [22] and W-Net [23] approaches, since they are not capable of detecting the LC area automatically.…”

Frontiers in Retinal Image Processing

“…In the paper titled “Unsupervised Approaches for the Segmentation of Dry ARMD Lesions in Eye Fundus cSLO Images” [ 2 ], the authors propose an adaptation of a fully convolutional network, called W-net, as an efficient method for the segmentation of ARMD lesions. The model is tested on a large dataset and the results are shown to be promising.…”