Synopsis

The relationship between localized histological truth and various MRI quantitative measures were investigated using canine models. In this study, GRMD pectineus muscle samples were imaged on a 3T clinical scanner and a 4.7T small-bore scanner. Trichrome stained histology slice was registered to the 3D MRI volume. Localized quantitative MRI parameters are correlated with histology analysis (muscle, fibrosis and interstitial tissue percentage). Parameters derived from the 4.7T MRI data show consistent correlations, while those from 3T MRI data do not show consistent correlation.

Purpose

Methods

GRMD pectineus muscle samples were placed in 100ml centrifuge tubes filled with agarose gel to prevent any motion artifact. MRI imaging of the specimens were acquired on a clinical 3T Siemens Verio system and a 4.7T 40-cm bore Varian Bruker scanner. A quadrature bird-cage coil is used on the 4.7T acquisition and a surface coil is used on the 3T acquisition due to its superior SNR and small size of the muscle samples. The imaging parameters are shown in Table 1. On the 3T data, T1 map, T2 map, Dixon (water/fat) images, and DTI parameters (FA, TrD, and ADC) are provided by the scanner. After acquisition, T2-wighted, Dixon and DTI images were registered to T1-weighted images. All data from the 4.7T were processed offline. A non-local means filter was applied to all MRI images before quantitative parameter images were calculated¹. Despot1 approach with dual-flip angles was used to calculate T1 maps². T2 maps were computed using a linear fitting approach for the conventional spin echo images at 4 different echo times (20, 40, 60, 80ms). Water and Fat images were acquired using a 3-point Dixon method³. DTI parameters were calculated using Camino diffusion toolkit⁴ with weighted least squares algorithm. Water fraction (WF) and fat fraction (FF) values were calculated using the approach described in ⁵. After MRI imaging, muscle samples were sectioned at the middle and one 5-µm trichrome stained histology slice was generated per sample at 0.23 µm in-plane resolution. Trichrome images were segmented to determine composition of tissues (percentage of muscle, fibrosis, interstitial) with an in-house developed algorithm. Afterwards, histology images were registered to MRI volumes using the method described in ⁶. After registration, segmented histology images and MRI quantitative images were re-sampled to 0.39mm × 0.39mm resolution for the 3T data, and 0.24mm × 0.24mm resolution for the 4.7T data. The correlation is calculated by

$$\rho\left(m,h\right)=\frac{\tt E \it \left[\left(m-\mu_{m}\right)\left(h-\mu_{h}\right)\right]}{\sigma_{m}\sigma_{h}}$$

where m denotes the calculated MR parameter and h denotes the tissue composition based on histology, µ and σ denote mean and standard deviation, respectively.

Results

Figure 1 shows a set of MRI and histology images. Figure 2 shows a representative sample of registered MRI and histology slice. The fused images are shown in pseudo color. As shown, MR images are registered to histology images sufficiently to examine correlation. Figure 3 shows the correlation between muscle tissue percentage and MRI parameters. MRI quantitative parameters (FF, WF, FA, TrD, and ADC) from the 4.7T dataset correlate consistently with muscle percentages per samples. FF and FA parameters show negative correlation which can be explained with alteration of muscle cells with fat/fibrosis tissues. Positive correlation of WF and ADC values can be explained with increased water context in the muscle cells. However, quantitative MRI parameters from the 3T datasets do not show consistent correlation for all samples except the WF. Figure 4 is the same as Figure 3 except that the correlation is between MRI and fibrosis component percentage.

As shown, all quantitative MRI parameters obtained from the 4.7T data (except T1maps) show a significant correlation with fibrosis tissue percentage. Positive correlation of FF and WF values can be explained with increasing fibrosis cells along the fat cells. Similarly, negative correlation of ADC values can be interpreted based on decreasing water content with muscle cells. However, on the 3T data, fibrosis tissue percentage do not show coherent correlation for all samples.

Conclusion

Acknowledgements

References

1. Buades A, Coll B, Morel J-M. A non-local algorithm for image denoising. Proc of the 2005 IEEE Comp Soc Conf on Comp Vis and Pattern Recog. 2005; San Diego, USA.

2.Deoni SCL, Rutt BK, Peters TM. Rapid combined T1 and T2 mapping using gradient recalled acquisition in the steady state. Mag Res in Med. 2003; 49:515-526.

3. Glover GH, Schneider E. Three-point dixon technique for true water/fat decomposition with B0 inhomogeneity correction. Mag Res in Med. 1991; 18:371-383.

4. Cook PA, Bai Y, Nedjati-Gilani S, Seunarine KK, Hall MG, Parker GJ, Alexander DC. Camino: Open-Source Diffusion-MRI Reconstruction and Processing. 14th Sci Meeting of the Int Soc for Mag Res in Med. Seattle, WA, USA; 2759, 2006.

5. Reeder SB, Hines CD, Yu H, McKenzie CA, Brittain JH. On The Definition of Fat-Fraction for In Vivo Fat Quantification with Magnetic Resonance Imaging. Proc of Int Soc for Mag Res in Med; 2009.

6. Eresen A, McConnel S, Ji JX, Birch S, Kornegay S. MRI As a Diagnostic Imaging Alternative for DMD Therapeutic Assessments. 38th Ann Int Conf of the IEEE Eng in Med and Bio Soc; Orlando, FL, USA: IEEE; 2016.