Table 1 Literature survey.
Reference number | Technique | Dataset | Summary |
|---|---|---|---|
MHorUNet: High-order spatial interaction UNet | ISIC 2018 | MHorUNet leverages high-order spatial interactions for improved skin lesion segmentation, enhancing accuracy and robustness in complex skin lesion segmentation tasks | |
Ultralight VM-UNet with Parallel Vision Mamba | ISIC 2018 | Ultralight VM-UNet reduces the number of parameters significantly while maintaining performance, suitable for efficient skin lesion segmentation on resource-constrained devices | |
UNet with Attention-Guidance and Test Time Augmentation | ISIC 2017 | Attention mechanisms paired with UNet alongside test time augmentation result in better skin lesion segmentation performance when dealing with various test scenarios | |
DSEUNet: Lightweight UNet with Dynamic Space Grouping Enhancement | ISIC 2018 | DSEUNet is a lightweight model that employs dynamic space grouping to enhance segmentation, offering a balance between model complexity and segmentation accuracy | |
MSS-UNet: Multi-Spatial-Shift MLP-based UNet | PH2 + ISIC 2017 | MSS-UNet integrates Multi-Spatial-Shift MLP layers within the UNet architecture, enhancing spatial feature extraction for better skin lesion segmentation results | |
Triple UNet with Region of Interest Focus | ISIC 2018 | Triple UNet architecture focuses on regions of interest to improve skin lesion segmentation, particularly in complex and varied lesion environments | |
UNet and VGG Hybrid Architecture | ISIC 2017 | This technique combines UNet with VGG architecture for enhanced skin lesion analysis, utilizing the strengths of both models to improve segmentation and classification outcomes | |
Attention Residual U-Net with Improved Encoder-Decoder Architecture | ISIC 2018 | The attention residual U-Net enhances speed and accuracy of skin lesion segmentation through its encoder-decoder architecture that prioritizes important regions of interest | |
Improved U-Net with Enhanced U-shaped Network | PH2 | Improved U-shaped network for skin lesion segmentation, incorporating advanced techniques to enhance segmentation precision and efficiency | |
Improved U-Net with Contour Attention | ISIC 2018 | Enhanced traditional U-Net architecture by adding contour attention mechanisms, resulting in better segmentation accuracy, particularly for complex and irregularly shaped lesions | |
Improved Convolutional Neural Network (CNN) | ISIC 2017 | Improved CNN architecture to segment skin lesions, achieving enhanced accuracy by refining feature extraction and segmentation processes | |
Machine Learning with Image Segmentation | ISIC 2018 | Machine learning techniques combined with image segmentation for detecting and classifying skin diseases offer a systematic approach to improving diagnostic accuracy | |
Modified U-Net Architecture | PH2 + ISIC 2018 | Modified U-Net architecture tailored for skin lesion segmentation, enhancing feature extraction and improving segmentation performance | |
Convolutional Neural Network (CNN) | ISIC 2018 | CNN is utilized for skin lesion segmentation from dermoscopic images, focusing on improving the model’s ability to accurately detect and segment lesions | |
Fully Convolutional Networks (FCNs) | ISIC 2018 + PH2 | Comparison of different fully convolutional networks for skin lesion segmentation, providing insights into their effectiveness and performance in real-world scenarios | |
Fusion of U-Net and CNN | ISIC 2018 + PH2 | Fusion of U-Net with CNN models to enhance both segmentation and classification of skin lesions, offering a comprehensive approach to dermoscopic image analysis | |
MASDF-Net: Multi-Attention Codec Network with Selective and Dynamic Fusion | ISIC 2016, ISIC 2017 and ISIC 2018 | Proposes a deep learning approach for skin lesion segmentation, focusing on multi-attention mechanisms for better fusion and dynamic segmentation accuracy | |
Segmentation and Classification of Skin Lesions | Not specified | Discusses the application of segmentation and classification techniques for skin lesion detection and disease diagnosis | |
Deep Learning-based Skin Lesion Segmentation | ISIC 2018 | Utilizes deep learning models for skin lesion segmentation, demonstrating their effectiveness at an IEEE conference | |
Deep Extreme Cut (DXC) Method for Skin Lesion Boundary Segmentation | ISIC 2017 | Focuses on fully automated methods for skin lesion boundary segmentation using deep extreme cut methods | |
MultiResUNet Architecture for Multimodal Biomedical Image Segmentation | ISIC 2018 | Proposes a novel architecture, MultiResUNet, designed for improving segmentation accuracy in multimodal biomedical images | |
U-Net for Skin Lesion Segmentation | ISIC 2018 | Investigates the effectiveness of the U-Net architecture in skin lesion segmentation and explores optimal training strategies | |
Stochastic Region-Merging and Pixel-based Markov Random Field | PH2 | Proposes a hybrid method combining stochastic region merging and pixel-based Markov random fields for skin lesion segmentation | |
Transfer Learning Approach with U-Net and DCNN-SVM | PH2 dataset | Discusses a transfer learning-based skin lesion segmentation model combined with a deep convolutional neural network (DCNN) and Support Vector Machine (SVM) for melanoma detection | |
InSiNet: Deep Convolutional Approach | ISIC 2018, ISIC 2019, and ISIC 2020 | Introduces InSiNet, a deep convolutional network aimed at skin cancer detection and lesion segmentation, with a focus on improving accuracy | |
LSCS-Net: Lightweight Skin Cancer Segmentation Network | ISIC 2016, ISIC 2017, and ISIC 2018 | Proposes a lightweight skin cancer segmentation model, LSCS-Net, leveraging densely connected multi-rate atrous convolution for better performance | |
Attention Res-UNet with focal Tversky loss | ISIC 2017 ISIC 2018 | A deep learning model integrating attention mechanisms with residual U-Net for skin lesion segmentation. The model improves segmentation precision by focusing on lesion-specific regions. However, increased computational costs due to attention layers require optimization for practical deployment | |
MALUNet (Multi-Attention Lightweight U-Net) | ISIC 2018 | The lightweight U-Net architecture segments skin lesions while performing efficiently using attention mechanisms to improve feature specification. The model maintains good segmentation accuracy levels while being efficient in term of calculations yet demands additional development for broader skin condition and imaging modality applicability |
