Applying 3D texture algorithms on MRI to evaluate quality traits of loin

This study firstly proposed the use of 3D MRI images to analyze loins in a non-destructive way. For that, interpolation and reconstruction techniques are applied on 2D MRI images of loins and the computational texture algorithms were adapted to analyze the obtained 3D images. The influence of the i) MRI acquisition sequences (Spin Echo (SE), Gradient Echo (GE), Turbo 3D (T3D)), ii) 3D texture features algorithms (GLCM, NGLDM, GLRLM, GLCM + NGLDM + GLRLM), and iii) regression techniques (Multiple Linear Regression (MLR), Isotonic Regression (IR)) was also evaluated. Combinations of SE or GE with any texture algorithm and any regression technique gave accurate results, with correlation coefficients higher than 0.75 and mean absolute error lower than 2. However, considering not only the accuracy of the methodology but also the computational cost, the use of GE, GLCM and IR could be proposed to determine physico-chemical parameters of loins non-destructively.