FIGURE 3 - available via license: Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 International
Content may be subject to copyright.
Sample images of skin lesion with low boundaries contrast from ISIC-2017 dataset, including samples with hair and marks covering the lesion area, as well as complex samples with different skin color, lesion location, lesion size, lesion shape and so on.
Source publication
Accurate segmentation of lesion region from skin lesion images can provide guidance for medical experts to make an early diagnosis of skin cancer. In this study, we construct Recurrent Attentional Convolutional Networks (O-Net), which exploits the skin lesion’s attention class feature with a recurrent O-shape structure, in an iterative refinement s...
Context in source publication
Context 1
... Attention Class Feature Module: In general, the main challenge of skin lesion images segmentation is to distinguish the low contrast voxels in the boundary regions. Figure 3 shows the challenging skin lesion images from ISIC-2017 dataset. Traditional non-learning-based approaches guide the lesion segmentation process through obtain spatial color distribution of the lesion image, which may bring bias because of the lowest tissue contrast arise from the edge region. ...
Similar publications
![The displacement distribution along x,y\documentclass[12pt]{minimal}...](publication/368716003/figure/fig5/AS:11431281255927164@1719455451807/The-displacement-distribution-along-x-ydocumentclass12ptminimal-usepackageamsmath_Q320.jpg)
The forced vibration of rectangular thin plate with a pair of edges simply supported and another pair of edges clamped mounted on the fractional standard solid foundation under the standing and moving loads is studied in this paper. Compared with the traditional integer-order viscoelastic foundation model, the fractional-order viscoelastic foundati...
Citations
... The model's effectiveness lies in its ability to maintain a seamless flow of information between the encoder and decoder, enabling precise delineation of lesions in the challenging scenarios posed by the ISIC dataset [143,144]. The incorporation of skip connections proves pivotal in retaining fine-grained details during upsampling, a crucial factor contributing to U-Net's remarkable success in skin cancer segmentation [145,146]. The following equation can succinctly represent the architecture of U-Net: Output = σ (Decoder (Encoder(Input) + Skip Connections)) (11) where σ denotes the sigmoid activation function, and the skip connections play a crucial role in preserving detailed information during the segmentation process. ...
The International Skin Imaging Collaboration (ISIC) datasets are pivotal resources for researchers in machine learning for medical image analysis, especially in skin cancer detection. These datasets contain tens of thousands of dermoscopic photographs, each accompanied by gold-standard lesion diagnosis metadata. Annual challenges associated with ISIC datasets have spurred significant advancements, with research papers reporting metrics surpassing those of human experts. Skin cancers are categorized into melanoma and non-melanoma types, with melanoma posing a greater threat due to its rapid potential for metastasis if left untreated. This paper aims to address challenges in skin cancer detection via visual inspection and manual examination of skin lesion images, processes historically known for their laboriousness. Despite notable advancements in machine learning and deep learning models, persistent challenges remain, largely due to the intricate nature of skin lesion images. We review research on convolutional neural networks (CNNs) in skin cancer classification and segmentation, identifying issues like data duplication and augmentation problems. We explore the efficacy of Vision Transformers (ViTs) in overcoming these challenges within ISIC dataset processing. ViTs leverage their capabilities to capture both global and local relationships within images, reducing data duplication and enhancing model generalization. Additionally, ViTs alleviate augmentation issues by effectively leveraging original data. Through a thorough examination of ViT-based methodologies, we illustrate their pivotal role in enhancing ISIC image classification and segmentation. This study offers valuable insights for researchers and practitioners looking to utilize ViTs for improved analysis of dermatological images. Furthermore, this paper emphasizes the crucial role of mathematical and computational modeling processes in advancing skin cancer detection methodologies, highlighting their significance in improving algorithmic performance and interpretability.
... When contrasted with conventional convolutional neural networks (CNN), architectures that integrate the strengths of both convolutional neural networks and recurrent neural networks (RNN), such as UNet, have attracted more and more attention because they allow form detailed representation. Moreover, recent studies have highlighted the efficacy of enhancing performance through the incorporation of an attention block, enabling the exploitation of a more comprehensive context [5]. ...
... This network was evaluated on three public datasets: ISIC 2017, ISIC 2018, and PH 2 . In parallel, a study by P. Chen et al. [5], introduced a similar approach, the O-Net, this network uses a recurrent O-shaped structure inspired by R-UNet (Recurrent UNet) and ACFNet (Attentional Class Feature Network). The system incorporates a module for attention feature fusion within the recurrent unit, which can gradually refine the segmentation considering the attention information of the lesion and capture sufficient contextual information. ...
... This dataset employs RGB JPEG images. Image resolutions range from 566 x 679 pixels to 4499 x 6748 pixels, and corresponding binary lesion annotations are provided in PNG format, maintaining consistent resolution with their respective images [5]. The test data consists of 600 images. ...
The challenges posed by skin diseases, including
cancers have led to growing interest in automating
dermatological diagnostics using machine vision and deep
learning. Models such as UNet and its derivatives have shown
their effectiveness in the segmentation of skin lesions. In this
article, we introduce an automated deep-learning system
designed to segment and classify skin lesions hierarchically,
while providing the user with information useful for
interpreting the classification results. Our study focuses first on
evaluating the performance of a combined Attention module
with UNet model «Att-UNet» to detect pertinent regions within
skin lesions against baseline models, and makes classification
decisions using different networks such as CNN, VGG16 and
Resnet50. Evaluated on the ISIC17 and ISIC16 databases, the
results highlight the specific advantages of each model, in
particularly the positive impact of the integration of attention
mechanisms. All models exceeded a threshold of 90% precision,
with a Dice coefficient of 0.953 on the validation set and 0.966
for the test set for Attention-UNet.
... The essential process of detecting the borders of skin cancer lesions is termed as segmentation. Accurate segmentation of skin lesion in medical imaging is one of the important research direction in modern medical era and it is a key step towards diagnosis and treatment of skin cancer [4,5]. However, wide range of lesion sizes, colours, textures, irregular forms and shapes of healthy skin, as well as detecting occlusions brought on by hair, blood vessels, or other skin features, precise automatic segmentation can be difficult, as shown in Figure 1. ...
Skin cancer is common and deadly, needs to be detected and treated properly. Deep learning algorithms like UNet have shown potential results in medical imaging. Such approaches still struggle to capture fine‐grained details and scale differences in skin lesions‐based occlusions' appearance, size etc. This research proposes a redesign UNet, the Multi‐Scale Pyramid Attention Network (MSPAN), to improve skin cancer lesion segmentation. The input data is processed at numerous scales with varied receptive fields. This enhances the network's ability to identify lesion locations by capturing local and global context. Attention approaches also help the network to suppress noise by focusing on informative features. We have evaluated MSPAN model on the publicly available ISIC2018 benchmark dataset for skin lesion segmentation. The method surpasses traditional UNet and other current methods in accuracy and effectiveness. The model also has a post‐processing to estimate lesion area for fast inference, making it suitable for extensive screening. Redesigned UNet with the Multi‐Scale Pyramid Attention Network improves skin cancer lesion segmentation. The model's ability to collect fine‐grained information and handle occlusions allows for more accurate skin cancer diagnosis and treatment. The MSPAN design can improve computer‐aided diagnosis systems and help dermatologists make precise clinical decisions.
... Recurrent Attentional Convolutional Networks (O-Net) have been developed by Chen et al. [23]. O-Net was developed in order to extract data about attention feature maps and enable coarse-to-fine representations. ...
Skin cancer is a disorder that is becoming more prevalent around the world and is responsible for numerous mortality. Skin cancer starts in one organ and slowly moves to other parts of the body before killing the patient. Early detection of skin cancer is important for reducing the number of deaths all over the world. Because it takes time and money to manually diagnose skin cancer, it is critical to create automated diagnostic techniques to categorize skin lesions more accurately. Medical image enhancement and deep learning-based segmentation techniques are developed in this proposed work. Cancer affected and non-affected skin images are given as input for the proposed method. Data collection consists of raw data that cannot produce high accuracy. So, a certain pre-processing technique is used in the proposed method to achieve high accuracy. Dingo Optimized Texture based Histogram Equalization (DOTHE) strategy is utilized to improve the skin image. Then the pre-processed image is partitioned into different parts or regions according to the features and properties of the pixels in the image. U-Net network architecture is used in the proposed method to segment the enhanced image. The performances of the proposed model are analyzed using the Convolutional Neural Network (CNN) model. This proposed model is tested with several metrics which attain better performances like 97% accuracy, 96% sensitivity, 95% specificity, 94% precision, and 3% error. Thus the designed model enhances and segments the image effectively, and it is useful for effective skin cancer prediction.
... Recently, more variant networks have been proposed, mainly including recurrent neural networks [16][17], multiscale features [18][19], residual connections [20][21], attention mechanisms [22][23]. Among them, Zhang et al. [24] introduced an encoder-decoder architecture integrating multi-scale contextual information. ...
Accurate segmentation of organs and lesions from medical images holds paramount importance in aiding physicians with diagnosis and monitor diseases. At present, the widespread application of deep-learning in medical image segmentation is primarily attributed to its exceptional feature extraction capability. Nonetheless, due to blurred target boundary, wide range of changes and chaotic background, the segmentation of medical images is still faced with great challenges. To address these issues, we present a multi-level feature integration network (MFI-Net) with SE-Res2Conv encoder for jaw cyst segmentation. Specifically, we replace the original convolution operation with SE-Res2Conv to better maintain model’s capacity for extracting features across multiple scales. Then, a novel context extractor module including multi-scale pooling block (MPB) and position attention module (PAM), which aims to generate more discriminative features. Finally, a multi-level feature integration block (MFIB) is implemented within the decoder to efficiently integrate low-level detail features with high-level semantic features. Numerous experiments were conducted on both the original and augmented datasets of jaw cyst to demonstrate the advantages of MFI-Net, with results consistently superior to all competitors. The Dice, IoU and Jaccard values of our method reached 93.06%, 93.47%, 87.06% in the original database and 91.25%, 91.94%, 84.06% in the augmented database. Furthermore, the computational efficiency of MFI-Net is impressive, with a speed of 106 FPS at the input size of 3×256×256 on a NVIDIA RTX6000 graphics card.
... To aid dermatologists, numerous computer-aided techniques were presented in the literature. In addition, researchers introduced image processing techniques to analyse the lesion images and declare whether the image is of normal skin or cancerous skin [19]. All CAD systems consist of key steps involving image acquisition, border detection to separate lesion from healthy skin, feature extraction, and feature classification [20]. ...
In computer vision applications like surveillance and remote sensing, to mention a few, deep learning has had considerable success. Medical imaging still faces a number of difficulties, including intra‐class similarity, a scarcity of training data, and poor contrast skin lesions, notably in the case of skin cancer. An optimisation‐aided deep learning‐based system is proposed for accurate multi‐class skin lesion identification. The sequential procedures of the proposed system start with preprocessing and end with categorisation. The preprocessing step is where a hybrid contrast enhancement technique is initially proposed for lesion identification with healthy regions. Instead of flipping and rotating data, the outputs from the middle phases of the hybrid enhanced technique are employed for data augmentation in the next step. Next, two pre‐trained deep learning models, MobileNetV2 and NasNet Mobile, are trained using deep transfer learning on the upgraded enriched dataset. Later, a dual‐threshold serial approach is employed to obtain and combine the features of both models. The next step was the variance‐controlled Marine Predator methodology, which the authors proposed as a superior optimisation method. The top features from the fused feature vector are classified using machine learning classifiers. The experimental strategy provided enhanced accuracy of 94.4% using the publicly available dataset HAM10000. Additionally, the proposed framework is evaluated compared to current approaches, with remarkable results.
... Many authors in [26,29,36 [38]. It is notable that many recent works applied different attention mechanisms (channel, spatial, and/or atrous attention) in the CNN-based SLS methods [25,47,49,50,52,54,59,64,66,67,69,72,73,86,99,100,105,112,114,119,122,132,133,581,582]. Some authors also proposed semisupervised CNN-based SLS techniques [51,65,71,127]. ...
... Values Corresponding articles 0.25 [527] 0.90 [32,51,53,102,105,130,149,165,169,171,176,180,185,208,220,233,234,239,259,260,267,269,275,289,[298][299][300]305,351,362,365,383,384,434,440,460,468,510,512,519,523,529,595,607] 0.95 [146] Momentum 0.99 [119,138,159,205,271] 0.001 [233,260,491] 0.0001 [96,146,185,220,267,300,305,362,405,434,479,495,515,529] 0.00001 [82,146,351,463] 0.000001 [97,171] 0.00158 [198] 0.5 [236] 0.005 [53,160,275,407,512] 0.0005 [32,130,159,165,205,271,298] Decay rates 0.00005 [105,127,468] 0.50 [72,163,168,203,234,236] 0.60 [494] 1 0.90 [96,99,127,159,234,265,374,414,420,482,518] 0.990 [127,234,482,518] 0.995 [494] 2 0.999 [72,96,99,159,163,168,203,236,265,374,414,420] [59,68,69,95,114,129,138,141,150,173,177,179,182,184,198,202,217,230,237,245,250,257,259,261,269,272,278,285,307,369,392,398,401,413,426,607] TensorFlow (2015) C++, Python, CUDA 169K [68,97,182,184,198,202,209,218,222,224,230,257,275,277,280,283,285,286,294,303,307,369,397] [26,35,43,49,[51][52][53]55,[64][65][66][81][82][83][84]99,101,105,107,110,113,115,116,120,127,130,149,157,160,162,163,168,169,172,203,215,[218][219][220][221]224,232,235,236,240,271,288,289,359,362,379,380,383,391,396,405,407,409,421,428,576] a These values are the number of stars, taken from their official GitHub accounts [Access date: 07-Nov-2022]. ...
The Computer-aided Diagnosis or Detection (CAD) approach for skin lesion analysis is an emerging field of research that has the potential to alleviate the burden and cost of skin cancer screening. Researchers have recently indicated increasing interest in developing such CAD systems, with the intention of providing a user-friendly tool to dermatologists to reduce the challenges encountered or associated with manual inspection. This article aims to provide a comprehensive literature survey and review of a total of 594 publications (356 for skin lesion segmentation and 238 for skin lesion classification) published between 2011 and 2022. These articles are analyzed and summarized in a number of different ways to contribute vital information regarding the methods for the development of CAD systems. These ways include: relevant and essential definitions and theories, input data (dataset utilization, preprocessing, augmentations, and fixing imbalance problems), method configuration (techniques, architectures, module frameworks, and losses), training tactics (hyperparameter settings), and evaluation criteria. We intend to investigate a variety of performance-enhancing approaches, including ensemble and post-processing. We also discuss these dimensions to reveal their current trends based on utilization frequencies. In addition, we highlight the primary difficulties associated with evaluating skin lesion segmentation and classification systems using minimal datasets, as well as the potential solutions to these difficulties. Findings, recommendations, and trends are disclosed to inform future research on developing an automated and robust CAD system for skin lesion analysis.
... Figure 3 shows some of the images (512 × 512) used in the experiment. The ISIC 2017 dataset consists of 2750 dermoscopy images, and the BreCaHAD dataset has 162 breast cancer histopathology images which researchers widely use for the automatic detection and diagnosis of skin cancer and cancerous regions in breast cancer histology images [2,9,16,18,27]. The USC-SIPI database contains popular standards images, namely Lena, Mandrill, Airplane, etc., widely used in digital image processing and salt and pepper noise removal algorithms. The optimal threshold (Tol) is to be calculated to give the best restoration result. ...
A new method for eliminating salt and pepper noise from color medical images is formulated in this work. The presence of noise in the medical images degrades image quality, affecting disease analysis, detection, and diagnosis by the doctors. Therefore, removal of noise from the medical image is crucial. For color image, vector median filter is preferred for decreasing presence of salt and pepper noise as it preserves the correlation between the channels. However, applying filter on the image without detecting the noise not only reduces noise, but also produces blurring effect in the homogeneous regions and removes the important features such as textures, edges, thin lines, curves, corners etc. presence in the images. This paper proposes a switching vector median filter that detects the salt and pepper noise in the images prior to the filtering operation to avoid such undesirable effects. The vector median filter is applied in the filtering kernel if the central vector pixel does not lie in the set of healthy vector pixels and the minimum average sums of the distances of the vector pixels that forms the edges in the four directions is more than a predetermined threshold. In comparison to existing common filters, the simulation results demonstrate the proposed filter’s superior performance for color medical image in decreasing salt and pepper noise and maintaining details.
Dermatosis are prevalent across different age groups, and using deep learning methods to assist general practitioners can improve the accuracy of their diagnoses. This paper summarizes the applications of deep learning in the field of image processing, particularly in the segmentation and classification of skin disease images. First, it introduces the main deep learning models used in image segmentation and classification. Then, it provides a detailed overview of the specific applications and improvements of various segmentation and classification models in the task of skin disease image processing. By summarizing relevant studies, it demonstrates the significant advancements in accuracy achieved by deep learning in skin disease image processing. Finally, the paper concludes with a summary and offers prospects for the future of intelligent skin disease diagnosis.
