Improved MR-based characterization of engineered cartilage using multiexponential T2 relaxation and multivariate analysis

Abstract

Noninvasive monitoring of tissue quality would be of substantial use in the development of cartilage tissue engineering strategies. Conventional MR parameters provide noninvasive measures of biophysical tissue properties and are sensitive to changes in matrix development, but do not clearly distinguish between groups with different levels of matrix development. Furthermore, MR outcomes are nonspecific, with particular changes in matrix components resulting in changes in multiple MR parameters. To address these limitations, we present two new approaches for the evaluation of tissue engineered constructs using MR, and apply them to immature and mature engineered cartilage after 1 and 5 weeks of development, respectively. First, we applied multiexponential T(2) analysis for the quantification of matrix macromolecule-associated water compartments. Second, we applied multivariate support vector machine analysis using multiple MR parameters to improve detection of degree of matrix development. Monoexponential T(2) values decreased with maturation, but without further specificity. Much more specific information was provided by multiexponential analysis. The T(2) distribution in both immature and mature constructs was qualitatively comparable to that of native cartilage. The analysis showed that proteoglycan-bound water increased significantly during maturation, from a fraction of 0.05 ± 0.01 to 0.07 ± 0.01. Classification of samples based on individual MR parameters, T(1), T(2), k(m) or apparent diffusion coefficient, showed that the best classifiers were T(1) and k(m), with classification accuracies of 85% and 84%, respectively. Support vector machine analysis improved the accuracy to 98% using the combination (k(m), apparent diffusion coefficient). These approaches were validated using biochemical and Fourier transform infrared imaging spectroscopic analyses, which showed increased proteoglycan and collagen with maturation. In summary, multiexponential T(2) and multivariate support vector machine analyses provide improved sensitivity to changes in matrix development and specificity to matrix composition in tissue engineered cartilage. These approaches show substantial potential for the evaluation of engineered cartilage tissue and for extension to other tissue engineering constructs.

Figure 1

dGEMRIC derived fixed charge density (FCD) from chondrocyte-seeded hollow fiber bioreactors from (5). A) Mean FCD for control versus chondroitinase-ABC-treated cartilage. B) FCD versus biochemically-determined GAG content.

Figure 2

Mid-sagittal MR image of four engineered cartilage constructs threaded on a hollow tube, stacked vertically in one well of the four-well sample holder. The orange highlighted region represents the selected region of interest (ROI) of the top sample for MR parameter determination.

Figure 3

A) Representative infrared spectrum indicating specific spectral features for collagen and proteoglycan (PG). (B) Representative FT-IRIS image from half of an immature engineered cartilage construct showing the distributions of (i) PG and (ii) collagen normalized by biochemically-derived water content. (B,iii) Representative Alcian blue and H&E staining from the corresponding sample. The central void through which the sample was affixed onto the hollow support tube is seen as a semicircle along the bottom of the image. (C) Representative FT-IRIS image of an entire mature engineered cartilage showing the distribution of (i) PG and (ii) collagen normalized by biochemically derived water content. (C,iii) Representative Alcian blue and H&E staining from the corresponding sample. The central void for mounting onto the support tube is seen in the center of the image.

Figure 4

Typical T₂ distribution from a mature engineered cartilage sample showing components C_a, C_b, and C_c. The smaller peaks between C_a and C_b were detected in all samples to varying degrees but did not display magnetization fractions large enough for reliable quantification.

Figure 5

Component weight fractions for empty PGA scaffolds, immature and mature engineered cartilage. A) There are no differences in w_a between groups. B) w_b was not detected in PGA scaffolds and showed an increase between immature and mature groups. C) There was a decrease in the bulk water fraction, w_c, between PGA scaffolds and immature samples and a slight but insignificant decrease in w_c between immature and mature samples. Horizontal bars indicate p < 0.05.

Figure 6

A) Biochemically-derived PG content plotted against FT-IRIS-derived PG content normalized by water content (R² = 0.69). B) MR-derived PG water fraction w_b plotted against FT-IRIS-derived PG content normalized by water content (R² = 0.66).

Figure 7

Typical scatter plots for bivariate combinations of MR imaging parameters. Results for univariate classification based on simple means are shown along the top and right sides of each plot, respectively. ADC and km shows good separation between immature and mature sample groups. Diamond symbols: immature samples; Circle symbols: mature samples; Open symbols: training set; Solid symbols: validation set; Black solid symbols: misclassified samples in the validation set. The loci labeled −1 and +1 represent contours defined by the support vectors for immature and mature samples, respectively. The support vector margin is signified by the contour labeled zero.