Micron‐resolution Imaging of Cortical Bone under 14 T Ultrahigh Magnetic Field

Abstract Compact, mineralized cortical bone tissues are often concealed on magnetic resonance (MR) images. Recent development of MR instruments and pulse techniques has yielded significant advances in acquiring anatomical and physiological information from cortical bone despite its poor 1H signals. This work demonstrates the first MR research on cortical bones under an ultrahigh magnetic field of 14 T. The 1H signals of different mammalian species exhibit multi‐exponential decays of three characteristic T2 or T2* values: 0.1–0.5 ms, 1–4 ms, and 4–8 ms. Systematic sample comparisons attribute these T2/T2* value ranges to collagen‐bound water, pore water, and lipids, respectively. Ultrashort echo time (UTE) imaging under 14 T yielded spatial resolutions of 20–80 microns, which resolves the 3D anatomy of the Haversian canals. The T2* relaxation characteristics further allow spatial classifications of collagen, pore water and lipids in human specimens. The study achieves a record of the spatial resolution for MR imaging in bone and shows that ultrahigh‐field MR has the unique ability to differentiate the soft and organic compartments in bone tissues.


Image processing for MRI
Image processing and relaxation analyses were performed in MATLAB ® .The MIPAV package was used for the 3D visualization of the UTE-MRI images.For T 2 * relaxation analysis on the UTE-MRI, sections containing the cortical bone were selected.The region of interest (ROI) of the cortical tissue was determined on each slice (red ROI in Fig. S1b).
The mean signal of the water phantom (blue ROI) in each TE was used to calibrate the signal intensity of the ROI.For each voxel in the ROI, ILT or multi-exponential fittings were used to fit the T 2 * decay curves.Table S4 lists the fitting parameters.

Fitting methods for relaxation data
Typically, ILT was first used to find the position and distribution of T 2 * components for new datasets.ILT included a range of T 2 /T 2 * components from 10 -5 to 10 0 s and with 100 data points on a logarithmic scale.Single-exponential, bi-exponential, and tri-exponential fittings were also carried out by nonlinear least squares curve-fitting with the Levenberg-Marquardt algorithm and were tested under the Akaike information criterion (AIC) 66 : where  is the number of independently adjusted parameters within the model, and  is the maximum likelihood.When the errors in the model follow an independent normal distribution, and when  is the number of observations, and the sum square of residue (SSR) is the residual sum of squares, the AIC becomes: AIC measures the goodness of the data fitting.The most appropriate fitting corresponds to the fitting with the smallest AIC value.For the data of sheep bone in Fig. S2, the bi-exponential fitting yields the smallest AIC value.For the data of human bone in Fig. S2, the tri-exponential fitting yields the smallest AIC value.

Fig. S1 |
Fig. S1 | ROI selection examples.(a) original slice image on UTE-MRI.(b) ROI selection of water (blue) and cortical bone (red) for this slice.

Fig. S3 |
Fig. S3 | The T 2 * fitting results for the cortical femur bone of wild-type mice of 3, 6 and 12 weeks.The error bar represent the standard deviation of multiple bone samples.

Fig. S4 |
Fig. S4 | The T 2 * fitting results for the cortical femur bone of wild-type rats of 6 and 12 weeks.The error bar represent the standard deviation of multiple bone samples.

Fig. S5 |
Fig. S5 | Bi-exponential fittings of the spin-echo decays of the water signal of cortical bones studied in the MRS experiments.(a) rat, (b) sheep, (c) swine, and (d) bovine.

Fig
Fig. S8 | (a) Excised tissue from the femoral neck of an 80-year-old patient who underwent joint replacement surgery.The red box indicates the extracted specimen for the MR experiments.(b) UTE T 2 * decay of the human specimen at 14 T and the ILT fitting.(c) Direct excitation spectrum and (d) spin-echo spectrum (TE = 8 µs) of the human specimen.According to the deconvolution of the 1 H spectra, the overall lipid fraction of the human specimen was 9%.

Fig. S9 |
Fig. S9 | Ultrahigh-resolution MRI of human peri-cortical tissue with osteoarthritis (OA-2).(a) 2D 1 H intensity color map obtained from a UTE image at a resolution of 62.5 µm (FOV: 8×8×20 mm 3 , matrix size: 128×128×128).(b) T 2 * distribution of the whole cross-section.(c) Color map of the T 2 *-S fraction and (d) T 2 * distribution of the same region.(e) Color map of the T 2 *-M fraction and (f) T 2 * distribution of the same region.(g) Color map of the T 2 *-L fraction and (h) T 2 * distribution of the same region.

Fig. S10 |
Fig. S10 | Ultrahigh-resolution MRI of human peri-cortical tissue with osteoporosis (OP-1).(a) 2D 1 H intensity color map obtained from a UTE image at a resolution of 62.5 µm (FOV: 8×8×20 mm 3 , matrix size: 128×128×128).(b) T 2 * distribution of the whole cross-section.(c) Color map of the T 2 *-S fraction and (d) T 2 * distribution of the same region.(e) Color map of the T 2 *-M fraction and (f) T 2 * distribution of the same region.(g) Color map of the T 2 *-L fraction and (h) T 2 * distribution of the same region.

Fig. S11 |
Fig. S11 | Ultrahigh-resolution MRI of human peri-cortical tissue with osteoporosis (OP-2).(a) 2D 1 H intensity color map obtained from a UTE image at a resolution of 62.5 µm (FOV: 8×8×20 mm 3 , matrix size: 128×128×128).(b) T 2 * distribution of the whole cross-section.(c) Color map of the T 2 *-S fraction and (d) T 2 * distribution of the same region.(e) Color map of the T 2 *-M fraction and (f) T 2 * distribution of the same region.(g) Color map of the T 2 *-L fraction and (h) T 2 * distribution of the same region.

Fig. S16 |
Fig. S16 | The area fractions of different components in human bone samples