Automated Diagnosis of Skin Lesions using CNN and LSTM

In order to successfully treat skin lesions and enhance patient outcomes, early identification and correct diagnosis are essential. In this study, we describe a novel approach for the automated identification of skin lesions that combines long short-term memory (LSTM) networks with convolutional neural networks (CNNs). Our method leverages the power of CNNs for feature extraction from dermoscopy images, capturing both local and global information. These extracted features are then fed into an LSTM network to model temporal dependencies and learn the hierarchical structure of the data. The proposed architecture is designed to handle the inherent variability and complexity of skin lesion images, leading to a more accurate and robust diagnostic tool. When compared to existing algorithms, our method outperforms them in terms of sensitivity, specificity, and overall accuracy on a large dataset of annotated skin lesion images. Our findings show potential of proposed CNN-LSTM framework for automated skin lesion diagnosis, paving the way for more effective and accessible skin cancer screening tools.