DENOISING NONINVASIVE LASER-ULTRASOUND SIGNALS BASED ON COMBINATION OF WAVELET TRANSFORM AND INDEPENDENT COMPONENT ANALYSIS

Abstract

One of the medical applications of noninvasive laser-ultrasound is in the diagnosis of eye diseases. In such applications, specific features are detected in the received signals produced by generation and reflection of ultrasonic pulses in the intraocular interfaces. Due to the noisy nature of the process, the desired features are typically faded in the received signal. Therefore, denoising the signal is inevitable. The noise suppression technique, with the combination of wavelet transform and independent component analysis (WICA), is widely used for biomedical signals. However, when signals are not obtained by the multichannel recording systems, independent components (ICs) of the wavelet coefficients are not extracted completely by independent component analysis, and rejecting any extracted signals will cause data loss. This paper develops a new technique using improved version of WICA to eliminate this drawback for laser-ultrasound signals obtained from a single channel recording system. The approach is based on extracting ICs of wavelet detail coefficients of the noisy signal and applying the threshold on the ICs to reduce the noise. The proposed method is evaluated on two real laser-ultrasound signals by artificially adding white Gaussian noise to study the distortion measures of the filter outputs. The results of the study demonstrate superior performance compared with conventional denoising approaches such as WICA, wavelet denoising and median, Wiener and lowpass filters over a wide range of laser-ultrasound signal-to-noise ratios.