Answers are grounded in this paper and cite their sources.
Skin lesions caused by mpox may visually resemble those resulting from other skin diseases, making image-based classification a potential approach to support preliminary screening. This study aimed to compare the performance of a Convolutional Neural Network (CNN) and a Support Vector Machine (SVM) using combined Histogram of Oriented Gradients (HOG) and Local Binary Patterns (LBP) features for binary classification of mpox and other skin lesion images. The experimental dataset consisted of 224 unique original images, comprising 116 mpox images and 108 images of other conditions, including chickenpox and measles. The SVM employed a 23,250-dimensional HOG-LBP feature vector with a Radial Basis Function (RBF) kernel, whereas the CNN automatically learned visual representations from preprocessed RGB images. Both models were evaluated using identical stratified five-fold cross-validation partitions, with data splitting performed before augmentation and hyperparameter tuning conducted within the training partitions. Performance was assessed using accuracy, precision, recall, F1-score, and specificity. The CNN achieved an average accuracy of 84.37%, precision of 88.73%, recall of 80.25%, F1-score of 84.06%, and specificity of 88.92%. Meanwhile, the SVM achieved 83.03%, 84.87%, 81.92%, 83.32%, and 84.28%, respectively. The results indicate that the CNN achieved higher performance in four evaluation metrics, while the SVM obtained slightly higher recall. These findings demonstrate that both approaches offer different classification strengths within the evaluated dataset. Nevertheless, further evaluation using independent datasets and more diverse skin lesion images is necessary to assess their generalizability before potential application in preliminary mpox screening.
No takes yet. Share an insight, caveat, or question.
Saraswati et al. (2026) studied this question.
Synapse has enriched 10 related papers. Consider them for comparative context: