【摘要】 为探索从江香猪肌内脂肪含量的超声活体预测方法,该文选取110头从江香猪育肥猪开展活体测定,以超声图像为研究对象,利用Matlab R2015b软件提取灰度-梯度共生矩阵(gray gradient cooccurrence matrix,GGCM)及4个角度灰度共生矩阵(gray level cooccurrence matrix,GLCM)图像纹理特征参数。通过逐步回归分析法构建肌内脂肪(intramuscular fat,IMF)含量预测模型,再将此模型应用于另外42头从江香猪,验证模型准确性。逐步回归分析结果表明,背膘厚、灰度平均和梯度熵3个参数指标达到显著水平(P<0.05),回归预测模型决定系数R2=0.369。线性回归及相关分析得出,估测值与实测值间RMSE为0.686,皮尔逊积矩相关系数(pearson correlation coefficients)和斯皮尔曼相关系数(spearman correlation coefficients)分别为0.592和0.640(P<0.001)。该研究所构建的拟合回归模型可应用于从江香猪肌内脂肪活体预测,为后期高肌内脂肪含量从江香猪的选育提供更为便捷有效的手段。更多还原

【Abstract】 As one of the most important evaluating indicators of pork quality traits, the intramuscular fat(IMF) content has an influence on tenderness, flavor and succulence of the meat. Excessive and long-term selective breeding of high lean percentage and fast growth rate has resulted in generally lower IMF content in pigs. For traditional detection method, samples used for determination of IMF content usually come from slaughtering, which is difficult to operate and costs high. Congjiang pigs belongs to the characteristic local small breeds in China, and is well known for its small-sized body, slow growth, early sexual maturity, and low genetic diversity. The IMF percentage of Congjiang pigs reaches the domestic breeds standard level, but the population variation of IMF content is rather larger, with variation coefficient of 31.71%, suggesting efforts should be made to strengthen the breeding and improve group uniformity. The aim of the study was to predict the IMF percentage in longissimus muscle of live Congjiang pig using real-time ultrasound image. In this research, the body weight(BW), backfat thickness(BFT), loin muscle deepness(LMD) and two longitudinal real-time ultrasound images were collected across the 10 th to 11 th rib and 5 cm off-midline on live pigs from 110 Congjiang pigs. 31 candidate image parameters of gray gradient and 4 direction(0, 45, 90, 135 angle) graylevel cooccurrence matrix within a defined region(50*50 pixel region) located at the center of longissimus muscle across the 10 th to 11 th rib for each ultrasound image were obtained using image analysis software(Matlab R2015 b). After slaughter, a slice of longissimus muscle from left carcass across the 10 th to 11 th rib was cut off immediately for determining the IMF percentage by the petroleum ether extraction method. Each test was repeated three times, the mean value as the final IMF content. The model to predict longissimus muscle IMF percentage was developed using multivariate linear regression analysis with carcass longissimus muscle IMF percentage as dependent variables and BW, BFT, LMD and image parameters as independent variables. 42 Congjiang pigs were anew chosen for model validation by linear regression and correlation analysis of measured IMF and predicting IMF percentage. The result of regression analysis indicated that three independent variables which contained BFT and two image parameters of average gray(H6) and gradient entropy(H12) were significant in last model(P<0.05). The predictive equation is PIMF=6.443+0.064 BFT+0.031 H6–7.421 H12, with determinate coefficient R2 of 0.369. The determinate coefficient R2 between the predictive value obtained by model and measured value of the validation set was 0.381. The root mean square error between predictive value and measured value of the validation set was 0.686. Correlation analysis showed that the pearson correlation coefficients and spearman correlation coefficients were 0.592 and 0.640(P<0.001), respectively. Therefore, the regression model constructed in this study could be used to predict the living body of Congjiang pigs. Meanwhile, the experimental model provides a more convenient and effective nondestructive detection method for breeding with high IMF percentage of Congjiang pigs, which can reduce the breeding cost and shorten the breeding process, and promote the development of miniature pig’s characteristic industry.更多还原