Automated Monkeypox detection using deep learning: an inception-based approach

Monkeypox is emerging zoonotic disease that has attracted a great deal of attention due to its rising outbreak trends and similarity with other dermatological disorders, i.e., chickenpox and measles. Designing an automatic and stable detection system has been crucial to make early detection, evade misclassification, and prevent disease spread. Deep learning algorithms, particularly convolutional neural networks (CNNs), have come to be powerful algorithms in medical image processing. The current research proposes an Inception-based deep learning system to distinguish monkeypox lesions from other skin diseases. The dataset used in the work includes 5,064 images for monkeypox and 4,836 images for non-monkeypox diseases. The adopted model has 314 layers of which 94 are convolution layers and was trained for 25 epochs, achieving high accuracy of 96.71% along with a validation accuracy of 95.63%. The research in detail explains pre-processing techniques, training parameters, and hyper parameter tuning used to provide optimal performance. This also provides a comparison analysis with existing approaches, showing what advantages this model may give. The results show that deep learning is indeed possible for automating monkeypox detection with far greater accuracy and generalizability. The paper wraps up by discussing challenges and limitations, along with some suggestions for improvement for future research.