Enhancing Skin Cancer Diagnosis Using Swin Transformer with Hybrid Shifted Window-Based Multi-head Self-attention and SwiGLU-Based MLP

Deleted Journal, Journal Year: 2024, Volume and Issue: unknown

Published: June 5, 2024

Abstract Skin cancer is one of the most frequently occurring cancers worldwide, and early detection crucial for effective treatment. Dermatologists often face challenges such as heavy data demands, potential human errors, strict time limits, which can negatively affect diagnostic outcomes. Deep learning–based systems offer quick, accurate testing enhanced research capabilities, providing significant support to dermatologists. In this study, we Swin Transformer architecture by implementing hybrid shifted window-based multi-head self-attention (HSW-MSA) in place conventional (SW-MSA). This adjustment enables model more efficiently process areas skin overlap, capture finer details, manage long-range dependencies, while maintaining memory usage computational efficiency during training. Additionally, study replaces standard multi-layer perceptron (MLP) with a SwiGLU-based MLP, an upgraded version gated linear unit (GLU) module, achieve higher accuracy, faster training speeds, better parameter efficiency. The modified model-base was evaluated using publicly accessible ISIC 2019 dataset eight classes compared against popular convolutional neural networks (CNNs) cutting-edge vision transformer (ViT) models. exhaustive assessment on unseen test dataset, proposed Swin-Base demonstrated exceptional performance, achieving accuracy 89.36%, recall 85.13%, precision 88.22%, F1-score 86.65%, surpassing all previously reported deep learning models documented literature.

Language: Английский

---

Performance evaluation of E-VGG19 model: Enhancing real-time skin cancer detection and classification

Heliyon, Journal Year: 2024, Volume and Issue: 10(10), P. e31488 - e31488

Published: May 1, 2024

Skin cancer is a pervasive and potentially life-threatening disease. Early detection plays crucial role in improving patient outcomes. Machine learning (ML) techniques, particularly when combined with pre-trained deep models, have shown promise enhancing the accuracy of skin detection. In this paper, we enhanced VGG19 model max pooling dense layer for prediction cancer. Moreover, also explored models such as Visual Geometry Group 19 (VGG19), Residual Network 152 version 2 (ResNet152v2), Inception-Residual (InceptionResNetV2), Dense Convolutional 201 (DenseNet201), 50 (ResNet50), Inception 3 (InceptionV3), For training, lesions dataset used malignant benign cases. The extract features divide into two categories: benign. are then fed machine methods, including Linear Support Vector (SVM), k-Nearest Neighbors (KNN), Decision Tree (DT), Logistic Regression (LR) our results demonstrate that combining E-VGG19 traditional classifiers significantly improves overall classification classification. compared performance baseline metrics (recall, F1 score, precision, sensitivity, accuracy). experiment provide valuable insights effectiveness various accurate efficient This research contributes to ongoing efforts create automated technologies detecting can help healthcare professionals individuals identify potential cases at an early stage, ultimately leading more timely effective treatments.

Language: Английский

---

A robust deep learning framework for multiclass skin cancer classification

Scientific Reports, Journal Year: 2025, Volume and Issue: 15(1)

Published: Feb. 10, 2025

Skin cancer represents a significant global health concern, where early and precise diagnosis plays pivotal role in improving treatment efficacy patient survival rates. Nonetheless, the inherent visual similarities between benign malignant lesions pose substantial challenges to accurate classification. To overcome these obstacles, this study proposes an innovative hybrid deep learning model that combines ConvNeXtV2 blocks separable self-attention mechanisms, tailored enhance feature extraction optimize classification performance. The inclusion of initial two stages is driven by their ability effectively capture fine-grained local features subtle patterns, which are critical for distinguishing visually similar lesion types. Meanwhile, adoption later allows selectively prioritize diagnostically relevant regions while minimizing computational complexity, addressing inefficiencies often associated with traditional mechanisms. was comprehensively trained validated on ISIC 2019 dataset, includes eight distinct skin categories. Advanced methodologies such as data augmentation transfer were employed further robustness reliability. proposed architecture achieved exceptional performance metrics, 93.48% accuracy, 93.24% precision, 90.70% recall, 91.82% F1-score, outperforming over ten Convolutional Neural Network (CNN) based Vision Transformer (ViT) models tested under comparable conditions. Despite its robust performance, maintains compact design only 21.92 million parameters, making it highly efficient suitable deployment. Proposed Model demonstrates accuracy generalizability across diverse classes, establishing reliable framework clinical practice.

Language: Английский

---

Evolution of artificial intelligence as a modern technology in advanced cancer therapy

Journal of Drug Delivery Science and Technology, Journal Year: 2024, Volume and Issue: 98, P. 105892 - 105892

Published: June 15, 2024

Language: Английский

---

An Innovative Deep Learning Framework for Skin Cancer Detection Employing ConvNeXtV2 and Focal Self-Attention Mechanisms

Results in Engineering, Journal Year: 2024, Volume and Issue: unknown, P. 103692 - 103692

Published: Dec. 1, 2024

Language: Английский

---

An Improved Skin Lesion Classification Using a Hybrid Approach with Active Contour Snake Model and Lightweight Attention-Guided Capsule Networks

Diagnostics, Journal Year: 2024, Volume and Issue: 14(6), P. 636 - 636

Published: March 17, 2024

Skin cancer is a prevalent type of malignancy on global scale, and the early accurate diagnosis this condition utmost importance for survival patients. The clinical assessment cutaneous lesions crucial aspect medical practice, although it encounters several obstacles, such as prolonged waiting time misinterpretation. intricate nature skin lesions, coupled with variations in appearance texture, presents substantial barriers to classification. As such, skilled clinicians often struggle differentiate benign moles from malignant tumors images. Although deep learning-based approaches convolution neural networks have made significant improvements, their stability generalization continue experience difficulties, performance accurately delineating lesion borders, capturing refined spatial connections among features, using contextual information classification suboptimal. To address these limitations, we propose novel approach that combines snake models active contour (AC) segmentation, ResNet50 feature extraction, capsule network fusion lightweight attention mechanisms attain different channels regions within maps, enhance discrimination, improve accuracy. We employed stochastic gradient descent (SGD) optimization algorithm optimize model’s parameters. proposed model implemented publicly available datasets, namely, HAM10000 ISIC 2020. experimental results showed achieved an accuracy 98% AUC-ROC 97.3%, showcasing potential terms effective compared existing state-of-the-art (SOTA) approaches. These highlight our reshape automated dermatological provide helpful tool practitioners.

Language: Английский

---

Multi-functional Dressings for Recovery and Screenable Treatment of Wounds: A review

Heliyon, Journal Year: 2024, Volume and Issue: 11(1), P. e41465 - e41465

Published: Dec. 24, 2024

Language: Английский

---

Artificial Intelligence Diagnosing of Oral Lichen Planus: A Comparative Study

Bioengineering, Journal Year: 2024, Volume and Issue: 11(11), P. 1159 - 1159

Published: Nov. 18, 2024

Early diagnosis of oral lichen planus (OLP) is challenging, which traditionally dependent on clinical experience and subjective interpretation. Artificial intelligence (AI) technology has been widely applied in objective rapid diagnoses. In this study, we aim to investigate the potential AI OLP evaluate its effectiveness improving diagnostic accuracy accelerating decision making. A total 128 confirmed patients were included, lesion images from various anatomical sites collected. The was performed using platforms, including ChatGPT-4O, ChatGPT (Diagram-Date extension), Claude Opus, for directly identification pre-training identification. After feature training, platforms significantly improved, with overall recognition rates Opus increasing 59%, 68%, 15% 77%, 80%, 50%, respectively. Additionally, buccal mucosa reached 94%, 93%, 56%, However, less effectively when recognizing lesions common complex cases; instance, gums only 60%, 20%, demonstrating significant limitations. study highlights strengths limitations different technologies provides a reference future applications medicine.

Language: Английский

---

PMANet: Progressive multi-stage attention networks for skin disease classification

Image and Vision Computing, Journal Year: 2024, Volume and Issue: 149, P. 105166 - 105166

Published: July 4, 2024

Language: Английский

---

Avaliação do uso de inteligência artificial na detecção precoce de melanoma

Caderno Pedagógico, Journal Year: 2025, Volume and Issue: 22(1), P. e13330 - e13330

Published: Jan. 14, 2025

Introdução: A detecção precoce do melanoma é crucial para a redução da mortalidade associada essa forma agressiva de câncer pele, cuja incidência tem aumentado globalmente, com estimativas 324.635 novos casos e 57.043 mortes em 2020. Objetivo: Avaliar aplicação inteligência artificial (IA) na melanoma, explorando suas implicações clínicas, éticas sociais. Metodologia: Trata-se uma revisão narrativa literatura, abrangendo estudos publicados entre 2014 2024, português, inglês espanhol, que abordam relação IA melanoma. pesquisa foi realizada bases dados eletrônicas como PubMed, Scopus Web of Science, utilizando descritores controlados Medical Subject Headings (MeSH) Descritores Ciências Saúde (DeCS). Após triagem rigorosa, 16 artigos foram selecionados análise, considerando critérios relevância adequação à pergunta norteadora. Resultados: pode melhorar significativamente precisão diagnóstica, algoritmos demonstrando taxas sensibilidade superiores 90% alguns estudos. No entanto, eficácia dos sistemas diretamente influenciada pela qualidade diversidade utilizados no treinamento, muitos conjuntos carecendo representatividade, especialmente tonalidade pele. Além disso, falta formação profissionais saúde as incertezas legais associadas ao uso dessas tecnologias emergem barreiras sua adoção. Conclusão: Ressalta-se necessidade colaboração interdisciplinar especialistas ciência computação, além importância diretrizes claras sobre responsabilidade legal. capacitação contínua essencial maximizar os benefícios prática clínica, garantindo implementação ética responsável possa, fato, contribuir e, consequentemente, melhoria desfechos saúde.

---