Impact of region of interest size on transcranial sonography based computer-aided diagnosis for Parkinson’s disease

Abstract

Transcranial sonography (TCS) has gained increasing application for diagnosis of Parkinson’s disease (PD) in clinical practice in recent years, because most PD patients, even in the early stage of PD, have abnormal hyperechogenicity of the substantia nigra (SN) in brainstem shown in TCS images. Therefore, the region of interest (ROI) for feature extraction should cover the SN region in a computer-aided diagnosis (CAD) system. The ROI size naturally affects the feature representation. However, there currently exist no unified standard for determining the size of ROI. In this work, we quantitatively compare the performance of TCS-based CAD with three sizes of ROIs, namely the entire midbrain (EM) region, the half of midbrain (HoM) region and the SN region. The experimental results on the original extracted features and the features by dimensionality reduction show that ROI covering the EM region achieves the overall best diagnosis performance. The results indicates that the neighboring regions around SN might also have abnormal symptoms, which cannot be clearly observed with naked eyes. It suggests that the large ROI includes more information for feature representation to improve the diagnosis performance of TCS-based CAD for PD.