Subcutaneous adipose tissue thickness determination using ultrasound signals processing: A phantom study. (August 2022)

• Record Type: Journal Article
• Authors: Hajiasgari, Mona; Setarehdan, Seyed Kamaledin; Rangraz, Parisa
• Is Part Of: Biomedical signal processing and control. Volume 77(2022)
• Publication Date: 2022-08
• Subjects: Ultrasound echo analysis -- RF data acquisition -- Subcutaneous adipose tissue characterization -- Wavelet transform -- Median filter -- Convert convex probe coordinates to linear; Signal processing -- Periodicals; Biomedical engineering -- Periodicals; Signal Processing, Computer-Assisted -- Periodicals; Image Processing, Computer-Assisted -- Periodicals; Biomedical Engineering -- Periodicals; 610.28
• DOI: 10.1016/j.bspc.2022.103744
• Languages: English
• ISSNs: 1746-8094

Abstract

Abstract: According to the top priority of health and fitness, the ultrasound (US) technique for measuring subcutaneous adipose tissue (SAT), as an indicator for body composition analysis, has received the most attention. So, tissue segmentation was performed to determine the boundaries between skin, SAT, and muscle in phantoms with different layers, and thicknesses that were fabricated to simulate tissue layers. The pre-processing of the signal was carried out using wavelet transform (WT) and Hilbert transform for denoising and detection. After the processing step, using WT for signal decomposition, a neural network was trained based on the scan line signals to detect the SAT layer. The coordinates of the convex probe were converted to linear to reduce the time and increase the accuracy for creating the target matrix. The algorithm was designed for automatic measurements of deep (normal), superficial (random), and noisy SAT thickness. Statistical evaluation was done to assess the system's skill in tissue classification and measurement of SAT thickness. Results revealed that most of the features related to coarse levels of detail coefficients extracted from wavelet decomposition levels can be used to build a classifier that can be applied successfully to differentiate between SAT and non-SAT tissue regions with a mean classification accuracy of 94.3% for 20 mm and 92.4% for 4 mm SAT in random mode with the error of estimate 0.05% and 0.07% respectively. Also, using 3 median filters and increasing their lengths from 3 to 7 improved the accuracy results to find SAT entry and exit boundaries.