A Novel MRI Compatible Soft Tissue Indentor and Fibre Bragg Grating Force Sensor

MRI is an ideal method for non-invasive soft tissue mechanical properties investigation. This requires mechanical excitation of the body’s tissues and measurement of the corresponding boundary conditions such as soft tissue deformation inside the MRI environment. However, this is technically difficult since load application and measurement of boundary conditions requires MRI compatible actuators and sensors. This paper describes a novel MRI compatible computer controlled soft tissue indentor and optical Fibre Brag Grating (FBG) force sensor. The high acquisition rate (100 Hz) force sensor was calibrated for forces up to 15 N and demonstrated a maximum error of 0.043 N. Performance and MRI compatibility of the devices was verified using indentation tests on a silicone gel phantom and the upper arm of a volunteer. The computer controlled indentor provided a highly repeatable tissue deformation. Since the indentor and force sensor are composed of nonferromagnetic materials, they are MRI compatible and no artefacts or temporal SNR reductions were observed. In a phantom study the mean and standard deviation of the temporal SNR levels without the indentor present were 500.18 and 207.08 respectively. With the indentor present the mean and standard deviation were 501.95 and 200.45 respectively. This computer controlled MRI compatible soft tissue indentation system with an integrated force sensor has a broad range of applications and will be used in the future for the non-invasive analysis of the mechanical properties of skeletal muscle tissue.


Introduction
Magnetic Resonance Imaging (MRI) is an ideal modality for the non-invasive analysis of soft tissue biomechanics as it provides excellent soft tissue contrast without exposing subjects to ionizing radiation. In addition it allows for the measurement of various biomechanical boundary conditions required for inverse analysis of tissue properties, such as 3D tissue geometry (segmentable from anatomical MRI), 3D architecture (e.g. based on diffusion tensor MRI [1]) and accurate 3D soft tissue deformation measurement (e.g. based on tagged MRI [2,3]). The non-invasive investigation of soft tissue mechanical properties allows for the validation of detailed constitutive laws enabling derivation of in-vivo tissue loading conditions and prediction of injury. Hence it has a wide array of applications including impact biomechanics [4], rehabilitation engineering [5] and surgical simulation [6].
However the MRI based investigation of the mechanical properties of soft tissue often requires an MRI compatible soft tissue loading system consisting of actuators (to mechanically palpate/excite the tissue) and sensor devices (to measure the applied load). Designing such devices to be safe within the MRI environment and compatible with the imaging is non-trivial [7]. Detailed evaluation of constitutive properties (including viscoelasticity) requires dynamic load application capabilities and, since imaging may require motion repetitions, displacement or force control and control of timing with respect to imaging. This paper describes a novel computer controlled MRI compatible soft tissue indentation system and force sensor enabling the MRI based analysis of the non-linear (visco)elastic and anisotropic behaviour of skeletal muscle tissue. The current focus is nonpainful and non-damaging indentation of the biceps region of the upper arm using a 45 mm diameter, which places a load constraint in the order of 15 N. Both quasi-static (load is held static during measurement of boundary conditions) and dynamic applications are of interest (non-painful indentation speeds whereby boundary conditions are continuously sampled as the tissue deforms).
Two main groups of actuator and sensor devices suitable for the MRI environment can be distinguished: 1) Devices employing electric principles and/or ferromagnetic components in the MRI room and/or close to the imaging region and 2) Devices which are intrinsically MRI compatible since they employ non-conducting and non-ferromagnetic materials and sensor signal transmission occurs using magnetically inert media within the MRI room. The latter group is the focus here since these do not significantly affect image quality and sensor performance. In addition they do not require shielding or anchoring and can safely be used in close proximity to the test subject. Therefore the applicability, safety and compatibility with MRI are more easily established for these devices which simplifies evaluation and approval (e.g. ethical) for a research setting.
A variety of soft tissue loading systems that can be used in the MRI environment have been proposed in the literature (e.g. [18,[22][23][24][25][26][27]), however their design is often tailored to suit their particular application (e.g. ranging from rat lower limb [18], to human footpad [25] and pigs buttock [23]) which, to date, has not included indentation of the upper arm. Despite this, a brief discussion of current soft tissue loading systems for the MRI environment is presented here. Most current soft tissue loading systems do not have dynamic displacement or force control capabilities during imaging (e.g. [22,24,25]) and have so far been limited to quasi-static applications (e.g. [18,[22][23][24][25][26]). In some cases the load can only be applied prior to imaging and cannot be altered during scanning (e.g. [22,24]) which does not enable dynamic analysis and MRI based deformation measurement techniques such as SPAMM (Spatial Modulation of the Magnetisation) (e.g. [2,3]) tagged MRI. Hence some researchers have employed contour tracking from matched anatomical MRI scans prior to and after loading (e.g. [24]) and registration techniques (e.g. [26]) or finite element analysis (FEA) for estimation of deformation instead (e.g. [21,22]). However these measures are more limited than full 3D deformation measurement based on SPAMM tagged MRI [3]. In addition the studies employing soft tissue loading devices have limited their analysis to the evaluation of isotropic hyperelastic constitutive models (e.g. [18,21,22,24,26,28,29]). Evaluation of more detailed constitutive models incorporating non-linearity, viscoelasticity and anisotropy require more detailed boundary condition measurement which is the focus of the current study.
Recently more advanced soft tissue loading devices have been proposed for the MRI environment such as Fu et al. 2011 [27] (a hydraulic actuator for lower leg indentation) and Solis et al. 2012 [23] (pigs buttock compression using a flat plate connected by a ~3 m rod (which passes through a wall) to a servo motor and force transducer outside the MRI room). Although both these systems feature dynamic displacement control capabilities and motion triggering toward the imaging, the latter can only operate approximately parallel to the MRI bore axis through the particular hole in the wall and for the former the displacement or force control and measurements achieved are not elaborated in sufficient detail.
For the soft tissue loading devices mentioned above the force measurement systems included the application of static weights [24], repetition of the experiment outside the MRI environment [22], the application of electric force sensors which suffer from electromagnetic interference [18] or force transducers outside the MRI room [23]. Other force sensors for the MRI environment have also been developed such as piezoelectric sensors. However these do not allow static force measurements and may induce image artefacts [12]. Since for the current study intrinsically MRI compatible devices are of interest optical force sensors are the focus. MRI compatible optical force sensors have been proposed for analysis of needle deflection and force feedback measurement during MRI based catheterisation have also been developed [15,30] however these are applied to relatively low forces in the range 0-0.5 N. Tokuno et al. [31]  However when in a later study [26] the authors used the system for human finger-tip indentation the force, acquired at 1 Hz, was found to vary with a standard deviation of up to 0.3 N around a static mean load of 2.35 N. The causes for these increased deviations, with respect to the calibration, were not discussed. Recently Song et al. [14] designed an advanced Fibre Bragg Grating (FBG) based triaxial force sensor mounted on a robot arm. Forces where calibrated up to ~10 N and a maximum force error of 0.5 N was recorded.
In this study a FBG sensor system is proposed because these have several advantages over other optical force sensing systems [14]: 1) measurement is independent of fluctuating light levels, 2) multiple gratings can be applied in series and 3) they have simple wiring and a compact implementation.
This paper outlines the design of a complete system for soft tissue indentation which is MRI compatible. This system is computer controlled and incorporates dynamic displacement control and motions can be triggered to be synchronised with the imaging (enabling the use of repeated image acquisitions common for MRI based deformation measurement techniques). Embedded in the MRI actuator is a novel high speed (100 Hz) MRI compatible optical FBG based force sensor enabling dynamic force measurement suitable for viscoelastic (e.g. ramp and hold type) analysis and fast detection of timing of motion and load application. The system is evaluated for the MRI based investigation of soft tissue biomechanics based on phantom tests and indentation of the upper arm of volunteers.

Fibre Bragg Grating Based Optical Force Sensing
In Fibre Bragg Grating (FBG) a periodic perturbation of the refractive index is introduced along an optical fibre acting as a local wavelength specific reflector [33,34]. The reflected (Bragg) wavelength for a specific grating is defined by [35]: Here is the effective refractive index of the fibre core and Λ is the period of the grating. From this equation it is clear that is both strain and temperature dependant since varies with temperature and Λ is altered following longitudinal fibre strain and thermal expansion/contraction [35]. Under isothermal conditions a linear relationship exists between reflected wavelength and the applied strain where tensile and compressive strains increase and decrease the wavelength reflected respectively. For the current study only the mechanical strain induced effect is of interest since it is linearly dependent on the force exerted on the optical fibre. In order to separate the effect of mechanical strain from the effects of temperature fluctuations two gratings are placed close together in series whereby one is subjected to both mechanical strain and local temperature variations, while the other is isolated from mechanical strain and acts as a (temperature) reference grating. The wavelength reflected from the latter reference grating is thus purely a function of temperature and, together with the wavelength reflected from the former (strain) grating , can be used to derive the mechanical fibre strain as [36]: Inc., USA). However for the current study the data was stored using a 10 point data interleave (running average of 10 consecutive data samples) resulting in an effective measurement frequency of 100Hz.
Due to the brittle nature of the fibre material (silica glass) it is best to load fibres in tension rather than compression. In tension the fibres used are capable of supporting loads up to 50N (corresponding to a 5% breakage strain). This is sufficient for the current study since measurement of forces up to 15 N are of interest (i.e. loads occurring during mild indentation of soft tissue).

The MRI compatible soft tissue indentor assembly
The design for the MRI compatible soft tissue indentor assembly will now be discussed with reference to FIG. 2. The MRI actuator body (1) is mounted on a support bridge (2) which is attached to two side plates (3) that are mounted onto the bottom plate (4). The whole assembly can be fixed on the scanner bed using a slide rail (5) and support ridge (6). The actuator orientation (blue arrows) can be adjusted using the adjustment screws (7) and (8) (1) allows water driven by the master cylinder to enter the actuator chamber formed via parts (2) and (3). The tube connector (4) and the rubber seal (5) form a manual valve which can be opened to allow for air removal during filling of the system. Parts (2) and (3)  which is inserted into the piston shaft (1). Part (7) of the force sensor assembly is fixed to the piston shaft via the screw (8). The FBG sensor fibre (9) enters the piston head at site (A) where it is supported using a bolt (10). The fibre reference and strain gratings (8 mm long each and 18 mm apart) are located at sites (B) and (C) respectively. The fibre (9) runs through a central hole in parts (6) and (7) where it is glued (EPO-TEK 353ND, Epoxy Technology Inc., USA) at sites (D) using the glue injection holes at (E). Thus when a compressive force FC is applied to the bottom of the indentor head it slides with respect to the piston shaft, converting the load to a tensile force FT at the strain grating (C) while the temperature reference grating (B) remains unloaded. The force FC is directly proportional to the force FT. Therefore due to the current design in the case of downward indentation with respect to gravity the weight of the indentor head components (0.94 N) first needs to be overcome. Hence for downward indentation only forces in excess of this weight can be recorded. Pre-tension can be

FIG. 3. The MRI actuator assembly in a retracted (left) and outward (right) state
introduced in the fibre using a screw (11) which also ensures that the weight of the indentor head assembly does not buckle the fibre.

Actuator motion control and data acquisition
The motion of the MRI compatible actuator discussed above is dictated by a hydraulic master cylinder assembly (see also FIG. 1C) placed outside the MRI room, which will be discussed using position. This provides a "map" of the indentor position for its full range of motion. The DC motor power and speed are adjusted through pulse width modulation of the input voltage. During the repeatable indentations the pulse width is ramped up to a set maximum (to achieve desired maximum motor power and speed), then as the indentor approaches its final position the pulse width is ramped down to a set minimum (for instance to achieve a desired minimum motor power during a hold phase) when the indentor approaches its set final position. Therefore in the current study a pulse width profile is prescribed controlling motor power and position (force or speed control can also be implemented).
Since the indentor location, orientation and final depth are set by hand prior to scanning (see  for the construction of FEA models derivable from anatomical MRI and 4) muscle tissue fibre architecture obtained from diffusion tensor MRI [1] to allow for analysis of anisotropic material behaviour.
In order to evaluate indentor system performance for the above application it was applied (FIG. 7) for indentation of a silicone gel phantom [37] and the upper arm region of volunteers (ethical approval and informed consent obtained from the Medical Ethical Committee, Academic Medical Centre, Amsterdam, The Netherlands). All the above mentioned boundary conditions were recorded during the experiments. However only the measurements relevant to the indentor system performance are highlighted in detail here as the soft tissue deformation measurements were presented elsewhere [2,3] and are beyond the scope of this paper. Since the quality of these measurements also illustrates the utility of the indentor system they are briefly summarized in the discussion section. Indentation cycles were triggered using a TTL pulse timed to start with the first dynamic or the first dynamic following completion of a deformation cycle.

Force measurement within the MRI environment
During the indentation experiments the FBG derived force was recorded at 100 Hz. Since the force sensor is optical fibre based no interference with the MRI imaging is expected. Since skeletal muscle tissue is highly viscoelastic [38,39] it is expected that, for the indentation tests of ramp and hold type discussed above, force relaxation should be observed during the hold phase. To demonstrate the sensor's performance in the MRI environment for application to soft tissue biomechanics, its ability to register this viscoelastic force history is demonstrated for two load rates of 10 mm/s and 20 mm/s.

Evaluation of indentor motion repeatability
In order to study repeatability of the indentor motion, the dynamic SPAMM tagged MRI series (see should therefore be observable in the data for the phantom and volunteer data respectively (e.g. for the phantom data dynamic 1 is repeated at dynamic 12 and 23, etc.). Analysis of parallel diagonal locations in the SSDM, showing difference magnitudes expected for noise (i.e. similar to differences between multiple static repetitions) thus allowed for demonstration of repeatability.

Evaluation of the system MRI compatibility
The MRI environment poses significant design challenges for the safe and appropriate functioning of both the device and the MRI scanner. In this study the following definition of MRI compatibility is used (for current definitions of MRI safety terminology see [7,40]. Although the term "MRI compatibility" is no longer favoured by the ASTM it is commonly used in the literature and hence also the device features presented, that are to be used in the MR environment are non-conducting (with the exception of the tap water used in the hydraulic system), non-metallic and non-magnetic the indentor system operation is not significantly affected by the MRI scanner and can be termed MR safe using scientific rationale [7]. In addition all materials employed in the MRI environment (e.g. polyoxymethylene, polyamide, polytetrafluorethylene and polyurethane) exhibit appropriate magnetic susceptibility [41] for MRI and thus it is likely that their influence on MRI data quality is minimal. Nonetheless to evaluate the effect of the indentor on MRI data quality, system performance was analysed in the MRI environment. MRI data was acquired for a silicone gel phantom (FIG. 7). In order to study the effect of the indentor presence on image quality a dynamic series (n=100) of MRI data was acquired with and without the indentor present. where ̅̅̅̅̅ and ̅̅̅̅̅ represents the mean and standard deviation of voxel ( , , ) respectively across all dynamics. After calibration based on the stair-case tests the response for a ramp test at 5 N/s was predicted (FIG. 9). The differences with respect to the load-cell force were not found to increase for the higher load rate which is also apparent from the large degree of overlap in FIG. 9. significantly altered by the presence of the indentor. The overall mean and standard deviation of the temporal SNR levels for the phantom volume without indentor present were 500. 18

Discussion
A novel MRI compatible soft tissue indentor system and optical FBG based force sensor have been presented. The computer controlled indentor motion is highly repeatable since MRI acquisitions during repeated indentor motions did not induce significant additional variation on top of what is expected due to noise. In addition the indentor device and force sensor are fully MRI compatible and are manufactured from non-ferromagnetic materials. The MRI compatibility was also evident following evaluation inside an MRI scanner and no negative effects such as SNR decrease and or image artefacts were observed.
Force measurement is sometimes based on the application of static weights [24] or by repeating the experiment outside the MRI environment [22] or electric force sensors which suffer from MRI scanning induced electromagnetic interference [18]. Recently Fu et al. [27] used an MRI compatible indentor system to study leg tissue biomechanics and 2D strain estimates were derived for 7 frames per indentation cycle based on harmonic phase MRI. However the actuator control and force sensing capabilities were not discussed in detail, hindering comparison to the current study. Solis et al. 2012 developed an advanced soft tissue loading system capable of motion control and triggering to MRI acquisitions. However the system (designed for porcine buttock compression) cannot be applied to  [43]) and for validation of motion compensation techniques for dynamic contrast enhanced imaging (FIG. 14B). The indentor has also found application outside the MRI environment for inverse mechanical property analysis combined with digital image correlation [44]. Other researchers have developed more complex tri-axial force sensor systems (e.g. [14,31,32,45]) for MRI. However the uniaxial force measurements presented here ensured a simple and compact indentor design and are sufficient for comparison to inverse FEA as the same resultant uniaxial force can easily be generated as an output.
For the current study the Fibre Bragg Grating signal was acquired at 1 kHz however a 10 point data interleave was used for the optical interrogator leading to an effective acquisition rate of 100 Hz. Although 100 Hz is deemed sufficient for the applications of the current study, a sample rate up to 1 kHz is feasible with the employed optical interrogator (or higher using more advanced interrogators). However acquiring the sensor data at 1kHz was not possible in the current study given the limitations in computer speed and of the indentor control and data acquisition software used.
The computer control system must simultaneously record the FBG signals and record and control the DC-motor behaviour and monitor the MRI TTL pulse in real-time. As such in the future force measurement of up to 1 kHz will be possible with this indentor given improvements in computational power and improvements in the data acquisition software.
With the exception of the optical interrogator system the proposed MRI compatible indentor system is relatively low cost. The MRI compatible indentor slave system components (used within MRI environment) are fabricated from non-conducting non-ferromagnetic materials which are all common engineering polymers (e.g. polyoxymethylene). All non-standard components can be manufactured using simple mill and rotating bench operations and all screws, tubes, fittings, the DCmotor and master cylinder are commercially available. If readers are interested in the design specifications or computer aided design files (based on Creo 1.0, Parametric Technology Company, MA, USA) these can be made freely available upon request.
Together with MRI modalities such as SPAMM tagged MRI, the indentor system allows for analysis of all boundary conditions required for the non-invasive investigation of the complex mechanical properties of soft tissue. Future work will focus on the use of the proposed system for the analysis of the non-linear elastic, anisotropic and viscoelastic mechanical properties of skeletal muscle tissue with application to the field of impact biomechanics and pressure ulcer prevention.
This is the first complete system for MRI based analysis of upper arm indentation in humans.

Conclusion
A novel MRI compatible indentor system is presented for the investigation of soft tissue biomechanics. A master slave system was developed whereby a computer controlled hydraulic master cylinder was used to provide highly repeatable motions to an MRI compatible actuator. To evaluate the system in the MRI environment and demonstrate its usefulness for soft tissue biomechanics investigation the system was used for indentation of a silicone gel phantom and the upper arm of volunteers. Repeatability was evident from the fact that MRI data for static repetitions showed similar variations as those from dynamic motion repetitions. All indentor components in the MRI room are non-ferromagnetic and non-conducting and hence MRI safe and can be used in close proximity to the imaging subject. MRI compatibility was demonstrated following imaging of a phantom and the presence of the MRI actuator did not induce any artefacts or significant SNR changes. Embedded in the indentor assembly is a novel high sampling speed (currently 100Hz) optical FBG based force sensor. The force sensor was calibrated for forces up to 15N and demonstrated a maximum force difference of 3.1% (maximum force difference magnitude was 0.043 N). Application of the force sensor in volunteer upper arm indentation showed the sensor's ability to register viscoelastic force decay resulting from ramp and hold indentation.