Search

Correlations between dry matter content and parameters determined by low-field nuclear magnetic resonance (NMR) relaxation and T1 weighted magnetic resonance imaging (MR-imaging) were investigated in four raw potato varieties using partial least-squares regression. The low-field NMR relaxation parameters T1(r=−0.85) and full Carr–Purcell–Meiboom–Gill (CPMG) relaxation curves (r=0.95) were correlated with the dry matter content of the potatoes. Furthermore, the amplitudes M21 and M22 of the two characteristic transverse relaxation (T2) time constants T21 and T22 obtained upon bi-exponential analysis of CPMG relaxation curves, were also highly correlated with the dry matter content (r=−0.66 and 0.78, respectively). MR-images showed very large visual differences between tissues, varieties and maturity. However, the applied image analysis features based on the histogram analysis were not correlated with dry matter content in whole potatoes in the present study. Thus, MR-imaging did not complement the correlations between NMR data and dry matter content. This means that the way the spatial distribution of water in the material was determined did not cover the full information about dry matter content. Therefore, the image analysis of MR-images requires more attention in the future. [Copyright &y& Elsevier]

The human ability to detect, interpret, assess and report visual stimuli is employed for qualitative and quantitative image analysis. Cross-sectional T1weighted MR images were acquired from a series of intact potatoes. Descriptive sensory image analysis was performed on the images. The sensory results were compared to conventional computer-assisted grey tone intensity histogram descriptors. A total of 60 tubers (2 varieties×2 storage times×15 replicate tubers) were submitted to MR-imaging. A trained panel of 9 assessors used 16 sensory descriptors to assess the images. These descriptors were developed by the panel during preliminary training sessions, and consisted in definitions of various biological compartments inside the tubers. The results from the sensory and the computer-assisted image analyses of the shape and interior structure of the tubers were related to the experimental design and to the dry matter content of the individual tubers, by partial least-squares regression (PLSR). Finally, predictive modelling of sensory image description from computer-assisted image analysis was attempted. The results showed that both the sensory and the computer-assisted image analyses were able to detect differences between varieties as well as storage times. The sensory image analysis gave better discrimination between varieties than the computer-assisted image analysis presently employed, and was easier to interpret. Some sensory descriptors could be predicted from the computer-assisted image analysis. The present results offer new information about using sensory analysis of MR-images not only for food science but also for medical applications for analysing MR and X-ray images and for training of personnel, such as radiologists and radiographers. [Copyright &y& Elsevier]