article · BioMedInformatics
Skin cancer is a widespread condition that can become life threatening without timely intervention. A dual-branch deep learning framework has been developed to improve skin lesion recognition by processing images through two parallel pathways. The first pathway pairs a lightweight convolutional neural network with bidirectional long short-term memory layers to capture local spatial and sequential patterns. The second pathway uses high-resolution vision transformers to extract broader global visual features. Extracted representations from both pathways are merged into a single classifier. To enhance performance and robustness, four distinct hybrid models incorporating different transformer architectures are combined using a logistic regression stacking technique. When evaluated on benchmark image collections, the stacked framework achieved 91.21 percent accuracy on a binary Kaggle dataset and 86.08 percent on the multi-class HAM10000 dataset, outperforming the individual hybrid models.
Accurate identification of skin lesions is essential for diagnosing skin cancer before it becomes life threatening. By combining complementary methods that assess both fine details and overall image context, this approach demonstrates how artificial intelligence can reliably classify complex medical imagery across both binary and multi-class diagnostic tasks.
The framework could support clinical decision-support tools used by dermatologists and general healthcare practitioners for automated skin lesion assessment. Because the findings are evaluated exclusively on public benchmark datasets, the technology remains at an early, algorithmic stage of development and would require extensive clinical validation before deployment in diagnostic medical software.
AI-generated from the published abstract. Always read the original work before citing.
Skin cancer is one of the most common diseases worldwide and, if left untreated, it can be life threatening. In this work, we propose a dual-branch deep learning framework that integrates a CNN–BiLSTM module for local spatial–sequential feature modeling with multiple transformer models (ViT, DeiT, SwinV2, and BEiT) for global representation learning. Each input image is processed in parallel by two branches. The first branch consists of a lightweight convolutional network, followed by row-wise and column-wise Bidirectional LSTM layers to capture spatial and sequential dependencies. The second branch employs a high-resolution transformer to extract features from the image. The features extracted from both branches of the hybrid CNN–BiLSTM–Transformer model are fused to construct a unified classifier for accurate skin lesion recognition. To further improve robustness, four CNN–BiLSTM–Transformer hybrid models are combined using a stacking strategy based on logistic regression that aggregates their prediction probabilities. Experimental results on the binary Kaggle and multi-class HAM10000 datasets demonstrate that stacking achieves better performance than the individual hybrid CNN–BiLSTM–Transformer models, reaching accuracies of 91.21% on the binary Kaggle dataset and 86.08 ± 0.47% on the HAM10000 dataset, respectively. These findings confirm the efficiency of using complementary local and global features for skin lesion classification.
This page summarises published work. The authoritative version sits with the publisher.
DOI: 10.3390/biomedinformatics6050066
Is something wrong with this record? Report it or request removal.
Discussion
Have you built on this work, tried to replicate it, or seen it applied in practice? Share what you know. Verified researchers and MARATTO™ domain experts can open a discussion, and any member can reply. Contributions are reviewed before they appear.
No discussion yet. Open the first thread.
New to MARATTO™? Create a free account.
