Magnetic resonance imaging

MRI is based on nuclear magnetic resonance. In the presence of a homogeneous magnetic field (B0), the nuclei of mainly ¹H or ²³Na will begin to precess about the direction ofB0 at Larmor frequency (ω0), and form the net magnetization. If a radio-frequency (RF) pulse is applied, the magnetization will tip away from the direction of the main magnetic field by a specific angle, called flip angle. Tipping the net magnetization also creates longitudinal and transverse vector components of the magnetization (Mzand Mxy, respectively).

After the application of the excitation pulse, the magnetization recovers (i.e relaxation), during which the nuclei exchange energy with each other (T2-, transverse- or spin-spin relaxation) or with their molecular surroundings (T1-, longitudinal- or spin-lattice relaxation) and emit echos. As the relaxation process depends on the chemical and physical environment of nuclei, the relaxation time is specific to individual tissues and fluids in the human body. MRI provides excellent soft tissue contrast. For this reason, various semi-quantitative and quantitative methods have been developed to classify knee joint OA based on MR images.

Semi-quantitative methods

There are three common semi-quantitative scoring methods: Whole-Organ Magnetic Resonance Imaging Score (WORMS) [22], Boston-Leeds Osteoarthritis Knee Score (BLOKS) [25] and MRI OsteoArthritis Knee Score (MOAKS) [152]. In each method, regional abnormalities in various soft tissues, such as cartilage, menisci and bones are assessed. Even though there are some differences in the definition and scoring of regions, all methods evaluate features present in OA: depth and extent of cartilage loss, cartilage defects, meniscal defects and subchondral bone marrow lesions.

Figure 3.2: Regional subdivisions of articular cartilage surfaces according to WORMS [22] system (left) and MOAKS [25] system (right) for the lateral joint compartment.

In the WORMS scoring system (seeFig. 3.1-left), the articular cartilage surfaces are divided as follows: femoral, tibial and patellar cartilage. For femoral and tibial cartilage, a further subdivision into anterior, central and posterior regions is done (Fig. 3.2-left). The patellar cartilage is divided into medial and lateral regions.

Then each region is scored using an eight-point scoring system. A similar regional subdivision is done in MOAKS (Fig. 3.2-right); however, a four-point scoring system is used. An illustrative example of different categories of WORMS, MOAKS and BLOKS scoring systems is shown inFig. 3.3.

Figure 3.3: Schematic diagram of the correspondence between categories from WORMS, MOAKS and BLOKS for describing articular cartilage loss [152]

In cohort studies with follow-up information, these methods may reveal longitudinal structural changes in the knee joint. Nonetheless, they are still susceptible to misclassification and inter- and intra-observer variability [29, 153].

and still offer insufficient information on the composition of soft-tissues.

Quantitative methods

T2 mapping. In the presence of an external magnetic field along the z-direction, a net magnetization forms in the z-direction, which is known as the longitudinal magnetization. When a RF-pulse is applied, this magnetization is tipped to the xy-plane or transverse xy-plane. TheT2 relaxation time (spin-spin relaxation ortransverse relaxation) is the process by which these transverse components of the magnetization decay. The process occurs at an exponential decay rate.

By acquiring several MR images at different echo times (TE) and then fitting an exponential signal decay function (S(TE)) to the signal intensity of each pixel, the T2relaxation time can be quantified:

S(TE)∝e^−TE/T2 (3.1) The signal intensity ofT2-weighted MR images varies with cartilage tissue depth [152, 154–159]. Although this behavior has been associated with fluid content or PG content [160], it is mainly indicative of the orientation of collagen fibrils in articular cartilage [154,161,162]. In healthy articular cartilage, typical values forT2relaxation times are ∼20-40ms in the superficial zone, ∼50ms in the middle zone and 10-20ms in the deep zone [163–167]. Therefore it is possible to evaluate the collagen architecture through depth-wise T2 relaxation time profiles throughout the tissue depth [162,166]. Importantly, local increases inT2relaxation times (seeFig. 3.1) have been associated with degeneration of the cartilage matrix, due to collagen fibrillation and increased fluid content [14, 24, 27, 156, 157, 168, 169]. Further,T₂relaxation times have also been associated with the mechanical properties of cartilage [167].

T1ρmapping. Relaxation process in a rotating frame of reference can be created by applying a continuous RF pulse or spin-lock pulse. This pulse locks the net magnetization in the transverse plane [31, 155, 157, 163, 170]. T_1ρ describes the longitudinal relaxation process while the spin-lock pulse is kepton. Similar to the T₂ relaxation time, the T_1ρ is directly proportional to the final image signal intensity. By applying several RF pulses at different spin-lock times (TSL) and then fitting an exponential signal decay function (S(TSL)) to the signal intensity of each pixel, theT_1ρrelaxation time can be quantified:

S(TSL)_∝e^−TSL/T1ρ (3.2) Even though the T_1ρ relaxation time carries some sensitivity to collagen [162, 171, 172], it is mainly indicative of PG content [27, 31, 156, 170, 173–176]. It has also been postulated thatT_1ρmay be indicative on general cartilage state, rather than any particular constituent [177]. In addition, theT_1ρ relaxation time is reportedly more sensitive to cartilage degeneration thanT2[27,178]. Moreover, theT2relaxation time is strongly dependent on the orientation of articular cartilage with respect to the main magnetic field [162].

Combined T₂ andT_1ρ mapping. Recently, an MRI sequence for simultaneous acquisition ofT₂andT_1ρrelaxation times was introduced [163]. Using this sequence, differences in articular cartilage [179], bone [180] and meniscus [26] between ACLR patients and healthy controls could be quantified. BothT2and T_1ρrelaxation times were significantly correlated with patient reported outcomes following ACLR [181]

and also matched structural changes quantified using WORMS grading [26, 179–

181].