Application of neural network in ultrasound tissue characterization using backscattered signal parameters

Real-time ultrasound (RTU) imaging methods for quality control of meats

“…The algorithms used to predict the IMF percentage of live animals were based on regression analysis (Whittaker et al ., 1992;Amin et al ., 1993;Newcom et al ., 2002;Li et al ., 2009); neural network (Brethour, 1994;Amin et al ., 1992;Harron and Dony, 2009;Li et al ., 2009) or support vector machine (Harron and Dony, 2009), among others. These algorithms were developed from textural RTU image features such as a histogram of pixel grey levels, Fourier-based (Amin et al ., 1997;Hassen et al ., 1999b;Newcom et al ., 2002;Harron and Dony, 2009).…”

“…It is also important to correctly distinguish the various tissue types -subcutaneous fat, muscle, blood capillaries, intramuscular fat and bones -for ROI box selection and subsequent use of computer image analysis (Amin et al ., 1997). During the image analysis process, attention must be paid to all these aspects since they affect the nature of the ultrasonic backscattered signal and consequently the quality of the image (Amin et al ., 1992). For each image, parameters were generated using texture analysis from a 100 × 100 pixels ROI box (Amin et al ., 1997;Hassen et al ., 2001;Newcom et al ., 2002).…”

Ultrasonic tissue characterization using texture analysis

“…The %fat values, using the ether extraction method, for various degrees of marbling were studied by Saveli et al (1986) and are Degree of Maturity (Haumschild and Carlson, 1983). Estimation of signal attenuation in Table 2.1: Mean ether extractable %fat of the rib eye classified according to their degree of marbling (Saveli et al, 1986) Degree the rib eye samples using the log-spectral difference method (Amin, 1989) shown good potential for evaluating marbling grades (Amin, 1992;Amin et al, 1992). Brethour (1990) reports the development of a scoring system for in-vivo estimation of marbling in live animals based on speckle present in B-mode images of the rib eye muscle.…”

“…Pattern recog¬ nition methods such as discriminant analysis and artificial neural networks can be used to develop a classification scheme based on the textural features for assigning quality grades. A feasibility study of applying neural networks for this application has already been done using features based on spectrum analysisof A-mode signals (Amin et al, 1992).…”