Magnetization transfer (MT) is a MR technique that generates contrast based on the exchange of magnetization between several groups of spins in different molecular environments. Both two-pool and three-pool models have been proposed to characterize the different groups of spins. The three-pool MT model divides the spins within a biological tissue into three groups: 1) a free pool A, composed of mobile protons; 2) a bound water pool B, composed of water protons bound to macromolecules and 3) a semisolid pool C, that consists of macromolecular protons¹. The two-pool MT model is highly simplified and only considers a water pool A and a macromolecule pool B. Theoretically, a three-pool model should be more accurate than the two-pool model for describing biological tissues.

Moreover, conventional MT modeling can only be applied to long T₂ tissues since short T₂ tissues such as cortical bone show litter or no signal with clinical sequences. Ultrashort echo time magnetization transfer (UTE-MT) imaging is likely to help with this difficulty. In this study we aimed to develop and utilize UTE-MT imaging and compare with two-pool and three-pool modeling of bovine cortical bone samples using a clinical 3T scanner.

Both two-pool and three-pool MT models have been described in details in the literature^1-4. Pool B is generally considered MR “invisible”. This is true with common clinical sequence but not correct with UTE sequences. With UTE-MT sequence, the signal equation is a combination of the steady-state longitudinal magnetization of pools A and B:$$S=M_z^Ae^{-TE/T_{2A}}+M_z^Be^{-TE/T_{2B}} [1]$$.

Where T_2A and T_2B are the T₂ value of pools A and B. TE is the echo time. Data with different TEs can be useful in separating pools A and B. In addition, the continuous wave power equivalent (CWPE) method⁴ for pulsed wave MT saturation used for the two-pool MT model can also be used for the three-pool modeling. Here, Gauss spectral absorption lineshape for the pool C was employed.

Data were acquired from a sectioned bovine cortical bone specimen (thickness=2cm) using a 2D non-slice selective UTE-MT sequence on a clinical 3T scanner (GE Healthcare Technologies, Milwaukee, WI). The UTE sequence employed a short rectangular pulse (duration=32µs) excitation followed by 2D radial ramp sampling with a minimal nominal TE of 8µs. The MT preparation utilized a Fermi shaped RF pulse (duration=8ms) and a gradient crusher. The UTE-MT imaging protocol included: TR=100ms, TE=8µs, FOV=4cm, matrix=128×128, five saturation powers (300°, 600°, 900°, 1200° and 1500°) and five frequency offsets (2, 5, 10, 20 and 50 kHz) with a total of 25 different MT dataset. In addition, UTE data with sixteen TEs (i.e. 0.01, 0.1, 0.2, 0.4, 0.6, 0.8, 1.0, 1.5, 2, 2.5, 3, 4, 5, 6, 7, 8 ms) were acquired for bi-component T2* analysis.

For data processing, two-pool MT modeling was employed first to provide useful information for further three-pool modeling, such as the T₂ value and fraction of semisolid pool C. These two parameters are fixed in the three-pool modeling in order to reduce the sensitivity to fitting errors. Finally the MT data and multiple TE data were combined together to fit Eq. [1].

Fig. 1 shows a representative UTE image of a bovine cortical bone sample, and the region of interest used for MT analysis. Fig. 2 shows the two-pool MT model and excellent fitting curves. Table 1 shows the fitting results. The macromolecule pool has a short T₂ of 14.5 µs and a fraction of 42.6%, consistent with results from NMR spectroscopy studies of cortical bone samples.

Fig. 3a shows the chain coupled three-pool model employed in this study. It is based on the assumption that the exchange rate between A and C is significantly less than the exchange rates between both A and B and B and C. The three-pool fitting curves are shown in the Fig. 3b and the corresponding fitting parameters are shown in Table 2. Pore water, bound water and macromolecule protons have T₂s of 1.63 ms, 0.27ms and 14.5µs, with fractions of 13.2%, 44.2% and 42.6%. These values are largely consistent with the literature.

We have demonstrated that both two-pool and three-pool MT modeling can be accomplished in bovine cortical bone samples with the UTE-MT sequence. The two-pool modeling and UTE bi-component analysis provide prior information useful for the three-pool modeling and reduces fitting errors minimized. The two-pool and three-pool UTE-MT modeling approach can be applied to many other short T₂ tissues such as menisci, ligaments, tendons, deep radial and calcified cartilage. It can provide a comprehensive evaluation of joint tissues degeneration in osteoarthritis (OA) and bone properties in osteoporosis (OP).