Wavelet K-Means Clustering and Fuzzy-Based Method for Segmenting MRI Images Depicting Parkinson's Disease

The early diagnosis of Parkinson's disease (PD) is important to reduce deaths due to this disease. Images of patients with PD obtained through magnetic resonance imaging (MRI) show gray matter, white matter, and cerebrospinal fluids; images of these components provide physicians with important information to determine the severity of the disease. Therefore, methods for segmenting these regions in images have been employed. However, difficulties have been encountered in the segmentation of the PD MRI images owing to the unclear boundaries between the gray matter and white matter and the regions being contained in homogeneous and unclear structures. Therefore, we propose a hybrid wavelet k-means clustering (KMC) and fuzzy median filter (FMF) method. First, detailed information from the MRI images was extracted using discrete wavelet transform; these images were enhanced by increasing the pixel values. The enhanced images were then fed into the KMC model for segmentation. Finally, the segmented images were input into the FMF for removing uncertainty and noise. This method can also be used for segmenting the MRI images of other diseases such as tuberous sclerosis (TBS) and atrial fibrillation (AF). The results of a qualitative and quantitative evaluation conducted using an open-source benchmark dataset and a clinical dataset were presented. The peak signal-to-noise ratio, structural similarity, and mean-squared error were used to compare the efficiency of the proposed method with that of other segmentation methods. The results proved that the proposed method can assist in the early detection of PD, TBS, and AF from MRI.