Enhancing Lyme Disease Erythema Migrans Rashes Classification: U-Net Segmentation and Ensemble Deep Learning Model

Tick-borne Lyme disease usually starts with Erythema Migrans rashes. Because of its visual complexity and resemblance to other skin disorders, Lyme disease is difficult to diagnose. It is crucial to promptly and accurately identify these skin rashes to ensure effective treatment and mitigate potential complications. This study proposes a distinctive approach that combines U-Net for image segmentation and an ensemble method based on deep learning to improve the robustness and classification accuracy in identifying Lyme disease-related rashes. Image normalization, data augmentation, and binary masking are all part of the preprocessing pipeline, along with the U-Net model applied for segmentation. Better performance is achieved by ensuring that the model concentrates on the most pertinent regions during this segmentation step. The ensemble model significantly enhances classification accuracy by utilizing a combination of InceptionV3, ResNet, and DenseNet. The ensemble achieved an accuracy of 99.50%, exceeding the accuracy of previous approaches. Thus, the study demonstrates that using our advanced ensemble deep learning models enhances the accuracy in classifying Lyme disease rashes, offering a more reliable approach for clinical diagnosis and improving patient care.