Dense Deconvolutional Network for Skin Lesion Segmentation

Abstract

Automatic delineation of skin lesion contours from dermoscopy images is a basic step in the process of diagnosis and treatment of skin lesions. However, it is a challenging task due to the high variations of appearances and sizes of skin lesions. In order to deal with such challenges, we propose a new dense deconvolutional network (DDN) for skin lesion segmentation based on residual learning. Specifically, the proposed network consists of dense deconvolutional layers (DDLs), chained residual pooling (CRP), and hierarchical supervision (HS). First, unlike traditional deconvolutional layers, DDLs are adopted to maintain the dimensions of the input and output images unchanged. The DDNs are trained in an end-to-end manner without the need of prior knowledge or complicated post-processing procedures. Second, the CRP aims to capture rich contextual background information and fuse multi-level features. By combining the local and global contextual information via multi-level feature fusion, the high-resolution prediction output is obtained. Third, HS is added to serve as an auxiliary loss and refine the prediction mask. The extensive experiments based on the public ISBI 2016 and 2017 skin lesion challenge datasets demonstrate the superior segmentation results of our proposed method over the state-of-the-art methods.

A Mutual Bootstrapping Model for Automated Skin Lesion Segmentation and Classification

“…We compared the proposed MB-DCNN model to several recently published skin lesion segmentation methods in Table I. On the ISIC-2017 dataset, the competing methods include a convolutional-deconvolutional neural network (CDNN) [8], a new dense deconvolutional network (DDN) [6], a fully convolutional network with star shape prior (FCN+SSP) [10], and a skin lesion segmentation deep model based on dilated residual and pyramid pooling network (SLSDeep) [9]. On the PH2 dataset, the competing methods consist of multi-stage FCN with parallel integration (mFCNPI) [7], a retrained FCN (RFCN) [4], and a simple linear iterative clustering (SLIC) method [11].…”

“…Yuan et al [4] developed a 19-layer deep FCN, which was optimized by using the Jaccard distance loss. Li et al [6] presented a new dense deconvolutional network based on the residual learning. Mirikharaji et al [10] proposed to encode the star shape prior into the loss function to guarantee a global structure in each segmentation result.…”

Fast ScanNet: Fast and Dense Analysis of Multi-Gigapixel Whole-Slide Images for Cancer Metastasis Detection

“…Recently, with the advance of computing power, deep learning methods have achieved remarkable success in many fields, including image classification [19]- [21], semantic segmentation [22]- [25], object detection [26], and natural language processing [27], etc. In the field of histology, Kumar et al released a dataset with ROIs extracted from whole-slide images and proposed boundary aware CNNs for nuclei segmentation [28].…”

Skin Lesion Segmentation Based on Multi-Scale Attention Convolutional Neural Network

“…The advantages and disadvantages of each method have been discussed and compared with many papers such as [11], [12], [23]. In recent years, with the continuous development of deep learning, the segmentation method based on CNN [14] was first applied in the field of image segmentation and achieved significant [7], [24]- [27] results in skin lesion segmentation.…”

“…In addition, a large number of artifacts including inherent skin characteristics (such as hair, blood vessels) and artificial artifacts (Such as air bubbles, ruler marks, uneven lighting, incomplete lesions, etc. ) make the task of skin lesion segmentation extremely difficult [9].…”