A Mutual Bootstrapping Model for Automated Skin Lesion Segmentation and Classification

Abstract

Automated skin lesion segmentation and classification are two most essential and related tasks in the computeraided diagnosis of skin cancer. Despite their prevalence, deep learning models are usually designed for only one task, ignoring the potential benefits in jointly performing both tasks. In this paper, we propose the mutual bootstrapping deep convolutional neural networks (MB-DCNN) model for simultaneous skin lesion segmentation and classification. This model consists of a coarse segmentation network (coarse-SN), a mask-guided classification network (mask-CN), and an enhanced segmentation network (enhanced-SN). On one hand, the coarse-SN generates coarse lesion masks that provide a prior bootstrapping for mask-CN to help it locate and classify skin lesions accurately. On the other hand, the lesion localization maps produced by mask-CN are then fed into enhancedSN, aiming to transfer the localization information learned by mask-CN to enhanced-SN for accurate lesion segmentation. In this way, both segmentation and classification networks mutually transfer knowledge between each other and facilitate each other in a bootstrapping way. Meanwhile, we also design a novel rank loss and jointly use it with the Dice loss in segmentation networks to address the issues caused by class imbalance and hard-easy pixel imbalance. We evaluate the proposed MB-DCNN model on the ISIC-2017 and PH2 datasets, and achieve a Jaccard index of 80.4% and 89.4% in skin lesion segmentation and an average AUC of 93.8% and 97.7% in skin lesion classification, which are superior to the performance of representative state-of-the-art skin lesion segmentation and classification methods. Our results suggest that it is possible to boost the performance of skin lesion segmentation and classification simultaneously via training a unified model to perform both tasks in a mutual bootstrapping way.

Melanoma Diagnosis Using Deep Learning and Fuzzy Logic

“…Our work was assessed on three well-established publicly available datasets PH2, ISBI 2017 Skin Lesion Challenge (SLC) and ISIC 2019 (SLC). We evaluated our proposed segmentation method against segmentation frameworks based on deep convolutional neural network (DCNN) [ 57 ], approaches with U-nets followed by histogram equalization and C-means clustering [ 58 ], segmentation done by crowdsourcing from ISIC 2017 challenge results [ 59 ], simultaneous segmentation and classification using bootstrapping deep convolutional neural network model [ 60 ], segmentation using contrast stretching and mean deviation [ 61 ] and semantic segmentation method for automatic segmentation [ 62 ]. In addition, we also drew inspiration from few of the most successful lesion segmentation methods introduced in the recent years like segmentation by means of FCN networks, multi stage fully convolution network (FCN) with parallel integration (mFCN-PI) [ 63 , 64 ], FrCN method involving simultaneous segmentation and classification, a fully-convolutional residual networks (FCRN), which was an amendment and extension of FCN architecture [ 65 , 66 , 67 ], a deep fully convolutional-deconvolutional neural network (CDNN) performing automatic segmentation [ 68 ] and lastly with the semi automatic Grab cut algorithm [ 69 ].…”

CDA-Net for Automatic Prostate Segmentation in MR Images

“…Compared with the binary cross-entropy loss function, this loss function can pay more attention to the segmentation of the foreground area, and can better deal with the problem of category imbalance. In recent years, it has gradually been widely used in many medical image segmentation competitions and papers [20,23,[46][47][48]. The DSC function is usually used to measure the overlap rate between the predicted segmentation and the ground truth, as shown in Equation (5).…”

ASCU-Net: Attention Gate, Spatial and Channel Attention U-Net for Skin Lesion Segmentation

“…However, this strategy for the detection of melanoma may be inaccurate or subjective, based on the experience of dermatologists alone [4]. In recent years, with the development of computer vision, medical image segmentation has become an important part of computer-aided diagnosis, which can support physicians in diagnosing dermoscopic images with speed and accuracy [5,6], providing professional interpretation of medical images [5]. However, segmentation of skin cancers is a challenging task because of the low image contrast and differences in color and size of skin lesions as well as the presence of air bubbles, hair, and ebony frames [7].…”