Publication

Abstract : A Genetic brain condition known as Huntington’s disease (HD) gradually destroys a person’s capacity for normal movement, thought, and behavior. Early detection of HD is crucial for a better lifestyle. EEG captures characteristic neural oscillation abnormalities in Huntington’s disease, reflecting disrupted cortical connectivity and progressive neurodegeneration. This work evaluates the use of the Daubechies wavelet to identify HD using electroencephalography (EEG) data. First, the EEG signals were filtered to remove noise and then segmented. Frequency domain features that include mean, energy, standard deviation, and entropy were obtained after applying different discrete wavelets like Symlet4(Sym4) and Daubechies(dB4). The sym4 and dB4 wavelets effectively capture transient, non-stationary features in EEG signals, such as spikes and rhythmic bursts, making them suitable for analyzing subtle neural variations in time-frequency space. Then, traditional machine learning classifiers that include decision tree, Random forest, SVM with RBF kernel, and Gradient Boosting were trained to differentiate between Huntington and healthy individuals. Grid search with cross-validation k=5 was used to find the best parameters for each model, and the decision tree classifier with dB4 wavelets provided better accuracy of 92.85% than Sym4 with 85.71% accuracy.