Burg power spectral density-based characterization of Doppler blood flow sound during hemorrhoidal artery ligation

Abstract

Hemorrhoidal artery ligation (HAL) has become universally accepted minimally invasive treatment of hemorrhoids disease. HAL involves precise identification of the superior rectal arteries supplying hemorrhoidal tissues using ultrasonic Doppler principles. During this process, at least there are three distinct sounds may be encountered by the surgeon. Only the pulsing Doppler sound is useful as it indicates the presence of hemorrhoidal artery. The accuracy based on traditional auscultation is commonly affected by surgeon’s hearing sensitivity and clinical experience. Therefore, automatic Doppler blood flow sound will be a great help in locating hemorrhoidal arteries. In this paper, a method based on the center frequency and kurtosis features extracted from Burg’s power spectral density (PSD) to distinguish three different types of Doppler blood flow sound signal during HAL procedure is proposed. Separability measurement was carried out using K– means clustering with the city block distance and three clusters corresponding to different sound types are successfully formed. In terms of arterial sound detection, an accuracy of 94.11% can be achieved. This result suggests that centre frequency, kurtosis, and maybe some other statistical features extracted from Burg PSD have the potential to be utilized as a means in automatic Doppler blood flow sound recognition.