JCV222: VISMED: EARLY SCREENING OF KAWASAKI DISEASE USING ARTIFICIAL INTELLIGENCE

ADZIIM MUHAIMIN BIN ROSDI MRSM PENGKALAN HULU

Background: Kawasaki disease (KD) can be fatal, as it inflames and swells blood vessels, potentially leading to heart complications in untreated children. It shares similar symptoms with the milder but contagious Hand-Foot-and-Mouth disease (HFMD), such as fever and rash. Problem: The early symptoms of KD and HFMD are almost identical. This may mislead parents into assuming HFMD instead of KD, potentially delaying vital treatment and risking life-threatening complications. Solution: The project developed an intelligent system using deep learning, utilising skin images and the AlexNet convolutional neural network (CNN) to detect and differentiate between KD and HFMD. Methodology: A dataset consisting of 707 image samples of KD, HFMD, and healthy controls was acquired from online databases. After preprocessing and augmentation, the dataset was randomly split for training and validation with 80:20 split ratio. Then, AlexNet was trained and validated using the acquired dataset. Outcomes: The network is successfully developed to classify skin images of KD, HFMD, and healthy control with accuracies of 99.8% for training, and 99.6% for validation. Potential: While the system is still in the prototype stage, the intelligent model can be enhanced to identify other diseases with symptoms emanating through the skin and miniaturised into a mobile application.