Early Localized Lyme Disease prediction using Image Processing and Machine Learning Techniques

The purpose of this study is to predict the early Lyme disease with the set of images and applying few machine learning algorithms. For this dataset taken from the public medical image repositories such as DermNet NZ, HAPiD, etc., using this two Convolutional Neural Network (CNN) machine learning algorithms such as VGGNet and GoogLeNet. The dataset size predicted using gpower tool with eighty percent of sample setting. With the sample size prediction the CNN type of VGGNet and GoogLeNet both are iterated and the results are recorded using training and testing data. The results were such as accuracy, precision, recall, F1-score, and ROC-AUC are determined to predict early localized Lyme disease. It was found that the novel CNN VGGNet model achieved an accuracy of 96.84% and other performance metrics compared to the CNN GoogLeNet algorithm’s 91.34% for the Lyme disease dataset. Hence this research study can help the healthcare practitioners or experts to identify the early growth of Lyme also this study result accepted the statistical significant of p value of 0.001, which is less than 0.05 in the independent t-test on both the machine learning algorithms. Therefore, the results indicate that the CNN VGGNet machine learning algorithm proves useful for early Lyme disease prediction and is effective at recognizing skin lesions.