T₂* MR Relaxometry and Ligament Volume are Associated with the Structural Properties of the Healing ACL

Abstract

Our objective was to develop a non-invasive magnetic resonance (MR) method to predict the structural properties of a healing anterior cruciate ligament (ACL) using volume and T₂* relaxation time. We also compared our T₂*-based structural property prediction model to a previous model utilizing signal intensity, an acquisition-dependent variable. Surgical ACL transection followed by no treatment (i.e., natural healing) or bio-enhanced ACL repair was performed in a porcine model. After 52 weeks of healing, high-resolution MR images of the ACL tissue were collected. From these images, ligament volumes and T₂* maps were established. The structural properties of the ligaments were determined via tensile testing. Using the T₂* histogram profile, each ligament voxel was binned based on its T₂* value into four discrete tissue sub-volumes defined by specific T₂* intervals. The linear combination of the ligament sub-volumes binned by T₂* value significantly predicted maximum load, yield load, and linear stiffness (R² = 0.92, 0.82, 0.88; p<0.001) and were similar to the previous signal intensity based method. In conclusion, the T₂* technique offers a highly predictive methodology that is a first step towards the development of a method that can be used to assess ligament healing across scanners, studies, and institutions.

INTRODUCTION

Biomechanical measurements of the structural properties of the anterior cruciate ligament (ACL) are frequently used to document functional healing after surgical ACL repair and reconstruction in pre-clinical animal models.^1–5 Despite being a useful quantitative measure of graft healing,^1–3 the current techniques to quantify the structural properties require harvesting the joint and testing the ligament to failure. Therefore, these current methods are inherently unsuitable for in vivo longitudinal assessment in both animal studies and human clinical trials.

Alternatively, magnetic resonance (MR) imaging is a widely available, non-invasive tool that has the potential to predict the biomechanical properties of ACL treatments.⁶ MR graft signal intensity has been found to correlate to the structural properties measured via ex vivo mechanical testing.³ Building on these initial studies, we found that the combination of MR ligament volume (a measure of tissue quantity) and the median ligament signal intensity (a surrogate measure of tissue quality) within that volume can be incorporated into a first order multiple regression model to improve the accuracy of the prediction of the structural properties.⁷ This new technique offered a more complete evaluation of graft integrity than either volume or signal intensity alone.⁷

However, the use of signal intensity as an outcome measure is limited by its dependence on image acquisition parameters and scanner manufacturer, rendering the predictions to be protocol, magnet, and hence, institution specific. One way to standardize MR results between scanners is to use relaxation time variables, such as T₂ and T₂*. These variables are inherent tissue properties that reflect specific tissue characteristics, and are much less sensitive to image acquisition parameters than conventional signal intensity data.⁸ T₂* relaxation time is a MR parameter that has been shown to correlate with the level of tissue organization, and is thus well suited for imaging highly organized collagenous structures,^9–12 such as ligaments and tendons. Thus, T₂* relaxation time could provide a more universal prediction model of the structural properties of a healing ligament that would be applicable across scanners of the same strength and between institutions.

The purpose of this study was to establish the relationship between ligament volume, T₂* relaxation time, and the structural properties of a healing ligament in a porcine model of ACL repair. We hypothesized that a multiple regression model based on ligament volume and its corresponding T₂* values would provide a noninvasive predictor of the ligament's structural properties after 52 weeks of healing. As a secondary aim, we compared the proposed T₂*-based structural properties prediction model to our previous model that incorporates signal intensity instead of T₂*.⁷ We hypothesized that the coefficients of determination would be greater and that the standard errors would be less when using the T₂* prediction method compared to the signal intensity method.

MATERIALS AND METHODS

Animal Model

Approval was obtained from the Institutional Animal Care and Use Committee prior to performing these studies. Fifteen adolescent Yucatan minipigs (approximately 15 weeks of age) underwent unilateral ACL transection surgery as previously described.^13,14 Immediately following transection, eight of the animals received bio-enhanced ACL repair with an extracellular matrix-blood composite (Bio-Enhanced Repair, or BE-ACL group) and seven were left untreated to heal naturally without repair (ACL transection, or ACLT group).¹³ All animals made it to 52 weeks with no complications, at which point all fifteen operative knees were harvested and immediately imaged. Following imaging, the specimens were frozen and stored at −20 degrees Celsius until mechanical testing.

MR Imaging

A surface knee coil on a 3T MR scanner (TIM Trio; Siemens, Erlangen, Germany) was used to image the joints. Two separate imaging protocols were performed on each knee: 1) a dual echo protocol to determine T₂* relaxation time, and 2) a manufacturer provided protocol to determine signal intensity.⁷ The images used for the signal intensity determination and analysis were a subset of those used in another study investigating the relationship between volume, signal intensity, and ligament structural properties over the course of healing.⁷ There was no intraarticular artifact found in any of the specimens. The BE-ACL group had a titanium button for suture fixation on the anterolateral cortical bone of the femur but was sufficiently far from the intra-articular space to avoid issues with artifact with the ACL.

MR Imaging: T₂* determination

To determine T₂*, a high-resolution T₁-weighted gradient echo 3-D FLASH dataset (note: T₁ weighted images are used to derive T₂* relaxation time) utilizing two echo times (TR/TE/FA, 25/7.36 & 15.24/ 12°; FOV, 140 mm; matrix 512×512; slice length/gap, 0.85mm/0; avg, 3; bandwidth, 130; scan time 19 minutes) was acquired of the injured knee immediately after harvest. High-resolution 3-D image acquisition was required to optimally capture the relatively small structure of the healing ACL. The healing ligaments and associated peri-ligamentous scar tissue were then manually segmented from these T₁-weighted MR images using commercially available software (Mimics 14.1; Materialize, Ann Arbor, MI). 3-D models of the healing ligaments were created using previously described methods.⁷ Summing the total number of ACL voxels provides an estimate of the whole ligament volume (16.1 voxels equaled one mm³). Using custom Matlab (R2012b; MathWorks, Natick, MA) code, T₂* maps were calculated using the signal intensity (SI) relationship: ${T_2}^{\ast }={\left(\frac{\ln S{I}_1-\ln S{I}_2}{T{E}_2-T{E}_1}\right)}^{-1}$,¹⁵ where SI₁ and SI₂ are the signal intensities corresponding to the echo times TE₁ and TE₂ where TE₂ > TE₁ for each voxel (note: TR was the same for both echo times allowing for the determination of T₂* using a two echo fit). To ensure the relaxation time maps used in this study were T₂* weighted, the images used to create the maps were gradient echo acquisitions, and the echo times were significant compared to the T₂* distribution of the tissue under examination.⁹

To produce ligament specific maps, the voxels corresponding to the ligament were extracted from the T₂* maps using the 3-D models created from the segmented T₁ weighted images (Fig. 1). Histograms of the voxel-wise T₂* values were plotted using these ligament specific maps. Two distinct peaks of relaxation times were apparent with no overlap within each healing ligament (Fig. 1). Further, the voxels making up these peaks were spatially organized such that the first voxel peak represented those with T₂*=0 ms and was generally located within the central portion of the ligament. Presumably, the voxels with T₂*=0 ms have a range of short T₂* values below our MR protocol's measurable limit (4.8 ms, the theoretical limit based on voxel signal to noise ratio)¹⁶, and would therefore fall between 0 and 4.8 ms. The second peak formed a lognormal distribution of relaxation times, where voxels with lower T₂* values were primarily found in the central portion of the ligament while higher T₂* values were identified towards the periphery (Fig. 1). The whole ligament volume was then binned into four separate tissue sub-volumes (Vol₁, Vol₂, Vol₃, Vol₄) with equal T₂* intervals up to 50 ms (0-12.5; 12.6-25; 25.1-37.5; 37.6-50 ms, respectively) (Fig. 2). Tissue volume in terms of mm³ (note: for this MR protocol 16.1 voxels equaled one mm³) was calculated for each sub-volume.

Figure 1.

Example ligament histogram showing (A) the bimodal distribution for T₂* with associated T₂* first quartile (Q1), median (Q2) and third quartile (Q3) summary statistics, (B) the T₂* ligament map, and (C) the original DICOM image. Note the ligament voxels illustrated in a red (B) represent voxels with a T₂* value of 0 ms. The MR images are a sagittal view of the femoral notch with the femur at the top of the image and the tibia at the bottom. For the MR images shown TE = 7.36 ms.

Figure 2.

T₂* model: (A) Actual versus predicted maximum load calculated using the linear combination of Vol₁, Vol₂, Vol₃ and Vol₄. The dotted lines represent the 95% confidence intervals. Gray shapes represent transected ligaments while black shapes represent repaired ligaments. The highest (star, B), median (square, C) and lowest (hexagon, D) maximum load ligaments and their corresponding histogram profile are also represented with associated T₂* first quartile (Q1), median (Q2) and third quartile (Q3) summary statistics.

MR Imaging: Signal intensity determination

During the same imaging session an additional single set of MR images was acquired to determine signal intensity. To accomplish this a T₂* weighted 3-D-CISS sequence (note: signal intensity was derived from a single set of T₂* weighted images) (TR/TE/FA, 12.9/6.5/ 35°; FOV, 140 mm; matrix 512×512, slice length/gap, 0.8mm/0; avg 1) was performed to establish the median signal intensity and volume of the whole ligament using our previously established multiple regression model.⁷ The healing ligaments were segmented from these MR images and 3-D models of the healing ligaments were created. From the models, the whole ligament volume (VW_SI) in terms of mm³ (note: for this MR protocol 17.1 voxels equaled one mm³) was determined. Histograms of signal intensity in terms of grayscale values normalized to the signal of posterior femoral cortical bone (normalization standard) were plotted.^9,17 Histograms of signal intensity were found to have a single uniform distribution for each ligament (Fig. 3). Signal intensity in terms of median gray scale value (MGV_SI) was calculated from each distribution for the whole ligament.

Figure 3.

T₂* model: (A) Actual versus predicted yield load (B) and actual versus predicted linear stiffness plots calculated using the linear combination of Vol₁, Vol₂, Vol₃ and Vol₄. The dotted lines represent the 95% confidence intervals.

Structural Properties of the healing ACL

An established tensile testing protocol was used to determine the structural properties of the repaired and untreated transected ACLs after 52 weeks of healing.^18,19 The specimens were thawed to room temperature. The proximal end of the femur and the distal end of the tibia were potted in PVC pipe using a urethane resin. The joint was carefully dissected, leaving only the femur-ligament-tibia complex intact and all associated peri-ligamentous scar tissue. Using a servohydraulic material testing system (MTS 810; Prairie Eden, MN), the tensile loads were applied at 20 mm/min to failure as previously reported.¹⁸ Initially, the joint was placed so that the mechanical axis of the ligament was collinear with the direction of pull of the actuator. Starting with a tibiofemoral compressive pre-load of 5 N, the entire tensile load-displacement curve was recorded until a precipitous drop in load occurred. The maximum load, yield load, and linear stiffness values of the ligaments were calculated from the load-displacement data.¹⁸

Data Analysis

First order multiple linear regression analyses (SigmaPlot 12.0; Systat Software Inc., San Jose, CA) were used to find the best-fit parameters and test the relationship between each ligament's sub-volume and the respective structural properties in the T₂* model. The resulting model included a volume term (Vol₁, Vol₂, Vol₃, Vol₄) representing each of the four bins. Each bin was defined by its associated interval of T₂* (0-12.5; 12.6-25; 25.1-37.5; 37.6-50 ms, respectively). Note that the voxels with a T₂* of 0 ms were included in Vol₁ because their relaxation times fall into the 0-12.5 ms bin range. The R-square values were reported as indicators of the relationship strength and goodness of fit. The p-values of the covariates (Vol₁, Vol₂, Vol₃, Vol₄) in the regression tested the contribution of the T₂* defined sub-volumes to the model. The predicted structural properties (maximum load, yield load, and linear stiffness) across specimens were plotted against the actual experimental structural properties to visualize the standard error of the regressions as an additional check of the T₂* model fit.

Additionally the volume of the whole ligament (VW_SI) and signal intensity (MGV_SI) values from each ligament were used in a first order multiple linear regression analysis to predict the structural properties as previously described.⁷ The predictions of the T₂* model and the signal intensity model were compared using the mean R-square values and their respective 95% confidence limits (CL).²⁰ Percent overlap of the confidence limits was tested using a z test statistic (alpha=0.05) to evaluate the differences between models.^21,22

RESULTS

T₂* Model Prediction

Of the T₂* derived parameters evaluated, the best prediction of mechanical properties was a linear combination of the ligament sub-volumes (Vol₁, Vol₂, Vol₃, Vol₄) defined by their respective intervals of T₂* values (Table 1). The R-squared values of the T₂* prediction equation for maximum load, yield load and linear stiffness were 0.93, 0.78, and 0.88, respectively (p<0.001 for all). The 95% confidence limits for these R-squared values of maximum load, yield load and linear stiffness were [0.92, 0.94]; [0.75, 0.81]; [0.86, 0.90], respectively. Standard errors for the prediction of maximum load (Fig. 2), yield load (Fig. 3A), and linear stiffness (Fig. 3B) were 109 N, 141 N and 26 N/mm respectively (Table 1).

Table 1.

T₂* model: Summary of ligament structural property prediction equations as a function of ligament sub-volumes (Vol₁, Vol₂, Vol₃, Vol₄) defined by range of T₂* values.

Dependent Variable (Y)	Predictor Term	T2* Range (ms)	P	R2	R2 95% CL	Standard Error	Predictor Term Contribution per unit volume (mm3)	Equati0n
Max Load (N)	Vol1	0 -12.5	<0.001	0.93	[0.92, 0.94]	109 N	1.55 N	Max Load = −161.5 + (1.55 × Vol1) + (1.20 × Vol2) - (1.62 × V0l3) - (5.48 × Vol4)
	Vol2	12.6 - 25	0.396				1.20 N
	Vol3	25.1 - 37.5	0.891				−1.62 N
	Vol4	37.6 - 50	0.746				−5.48 N
Yield Load (N)	Vol1	0 -12.5	0.005	0.78	[0.75, 0.81]	141 N	1.13 N	Yield Load = 55.3 + (1.13 × Vol1) + (0.04 × Vol2) - (1.24 × Vol3) - (1.64 x Vol4)
	Vol2	12.6 - 25	0.983				0.04 N
	Vol3	25.1 - 37.5	0.936				−1.24 N
	Vol4	37.6 - 50	0.940				−1.64 N
Stiffness (N/mm)	Vol1	0 -12.5	<0.001	0.88	[0.86, 0.90]	26 N/mm	0.28 N/mm	Stiffness = −1.8 + (0.28 × Vol1) + (0.02 × Vol2) + (0.55 × Vol3) - (0.70 × Vol4)
	Vol2	12.6 - 25	0.961				0.02 N/mm
	Vol3	25.1 - 37.5	0.845				0.55 N/mm
	Vol4	37.6 - 50	0.860				−0.70 N/mm

Signal Intensity Model Prediction

Of the signal intensity derived parameters studied, the best predictors of mechanical properties were a linear combination of the whole ligament volume (VW_SI) and the signal intensity (MGV_SI) (Table 2). The R-squared values of the signal intensity prediction equations for maximum load, yield load, and linear stiffness were 0.84, 0.92, and 0.88 respectively (p<0.001 for all) (Fig. 4). The 95% confidence limits for these R-squared values of maximum load, yield load and linear stiffness were [0.80, 0.88]; [0.90, 0.94]; [0.85, 0.91], respectively. Standard errors for the prediction of maximum load, yield load, and linear stiffness were 155 N, 75 N and 24 N/mm respectively (Table 2).

Table 2.

Signal Intensity model: Summary of ligament structural property prediction equations as a function of ligament whole volume (VW_SI) and signal intensity (MGV_si).

Dependent Variable (Y)	Predictor Term	p	R2	R2 95% CL	Standard Error	Equation
Max Load (N)	VWsI	<0.001	0.84	[0.80, 0.88]	155 N	Max Load = 299.8 + (0.8 × VWsI) - (424.7 × MGVsI)
	MGVsI	0.006
Yield Load (N)	VWsI	<0.001	0.92	[0.90, 0.94]	75 N	Yield Load = 340.2 + (0.5 × VWsI) - (372.4 × MGVsI)
	MGVsI	<0.001
Stiffness (N/mm)	VWsI	<0.001	0.88	[0.85, 0.91]	24 N/mm	Stiffness = 73.6 + (0.1 × VWsI) - (75.9 × MGVsI)
	MGVsI	0.002

Figure 4.

Signal intensity model: (A) Actual versus predicted maximum load calculated using linear combination of VW_SI and MGV_SI. The dotted lines represent the 95% confidence intervals. Gray shapes represent transected ligaments while black shapes represent repaired ligaments. The highest (star, B), median (square, C) and lowest (hexagon, D) maximum load ligaments and their corresponding histogram profile are also represented with associated SI first quartile (Q1), median (MGV_SI, Q2) and third quartile (Q3) summary statistics.

There was no overlap of the R-squared confidence limits between the T₂* and signal intensity models for the maximum load prediction. For maximum load the T₂* model displayed significantly higher R-squared confidence limits than the signal intensity model. There was also no overlap of R-squared confidence limits between the T₂* and signal intensity models for the yield load prediction. In this case the signal intensity model displayed significantly higher R-squared confidence limits than the T₂* model. There was a 100% overlap of R-squared confidence limits for linear stiffness with the T₂* prediction interval nested within the signal intensity model interval making them statistically equivalent (Tables 1 & 2).

DISCUSSION

A non-invasive tool that can predict the biomechanical properties of a healing ligament would be highly valuable in a research and clinical setting for evaluating outcomes of different ACL treatments. Our objective was to develop a magnetic resonance (MR) method to predict structural properties of a healing anterior cruciate ligament (ACL) using volume and T₂* relaxation time. We found the linear combination of the ligament sub-volumes defined by increasing T₂* intervals (Vol₁, Vol₂, Vol₃, Vol₄) significantly predicted structural properties of a healing porcine ACL at 52 weeks post-operatively.

There are two parameters that contribute to the structural properties of a ligament; 1) the amount of tissue (as represented by the volume) and 2) the quality of the tissue (as represented by T₂*). Using a T₂* histogram profile, each ligament was partitioned into four tissue sub-volumes with equal intervals of increasing T₂*. In our regression model, we determined the relative proportions of the total ligament volume made up of high quality tissue (Vol₁) down to the lowest quality tissue (Vol₄) to determine how these proportions contribute to the overall ligament structural properties. The amount of tissue with a T₂* value between 0-12.5 ms (Vol₁) was found to be the most significant in predicting structural properties (p<0.001; Table 1). This would be expected as highly organized tissue has been associated with short T₂* values¹⁰. Contribution of this sub-volume to structural properties can be observed with its associated slope in the prediction equations (Table 1). Per unit volume of Vol₁ (T₂* values 0-12.5 ms), 1.55 N, 1.13 N, and 0.28 N/mm are contributed to maximum load, yield load and linear stiffness, respectively. Generally less organized tissue in the T₂* range of 12.6-25 ms (Vol₂) contributed less to structural properties per unit volume with 1.20 N, 0.04 N and 0.02 N/mm for maximum load, yield load and linear stiffness respectively. While Vol₂, Vol₃ and Vol₄ were not found to be significant predictors in this model (p>0.3) they were kept as terms in the model to accommodate a broader range of T₂* values that would be expected with phases of healing earlier (<52 weeks) and to include the whole ligament volume in the analysis. The composition of the voxels in Vol₃ and Vol₄ with higher ranges of T₂* values is likely a combination of less densely packed collagen fibrils and periligamentous tissue of scar formation.

Additional analyses were performed to determine the effect of partitioning the ligaments into more sub-volumes (bins = 8) with smaller ranges of T₂* intervals defining each sub-volume (see Supplement 1). The four-bin model was the simplest linear regression model that yielded relatively high R-squared and low standard errors for determining structural properties when compared to other bin numbers.

The distribution of T₂* values for each ligament was found to be bimodal and exhibited two discrete peaks. The voxels in this first peak displayed a single T₂* value of 0 ms. Presumably there is variability in this tissue type, however the relaxation times approaching zero were too short to be acquired using our current MR protocol. The impact of this first peak is relatively small, however, as the median percentage of ACL voxels with T₂*=0 to the total number of voxels for the whole ACL was only 3.8% (1^st Quartile = 2.5%; 3^rd Quartile =8.3%). Nonetheless, we know that the zero T₂* values fall within the range of 0 and 4.8 ms. Despite not being able to capture the exact T₂* value of the voxels under the theoretical 4.8 ms limit¹⁶, the current approach of partitioning or binning each ligament into four sub-volumes based on the T₂* ranges helps to reduce the effects of minor observation error and allows these voxels to be accounted for in the prediction model. Unlike previous studies using a single average value for volume, signal intensity, or T₂;^7,23 quantifying sub-volumes (Vol₁, Vol₂, Vol₃, Vol₄) of tissue (with different levels of organization) based on their associated T₂* values offers a more detailed evaluation of tissue composition and its relationship to structural properties. Models that neglect the heterogenous composition of the healing ACL tissues do not account for the fact that these tissues likely have different material properties. Additionally, the histograms of individual ligaments (Fig. 2) can be observed for the contribution of the sub-volumes (Vol₁, Vol₂, Vol₃, Vol₄) to the structural properties prediction. The ligaments with the highest structural properties displayed larger sub-volumes of organized tissue represented by low T₂* times (Vol₁), while ligaments with the lowest structural properties displayed smaller sub-volumes of organized tissue (Fig. 2).

The T₂* model was found to be comparable to the original signal intensity prediction model. While the T₂* model did not display higher R-squared values for all structural properties as hypothesized, the 100% overlap seen in linear stiffness confidence limits between the T₂* and signal intensity methods, along with the marginal differences in maximum load and yield load confidence limits, suggest that these two prediction models offer similar certainty when determining structural properties.

However, the T₂* prediction model offers the benefit of using the T₂* variable, which is an inherent tissue property and much less sensitive to image acquisition parameters than conventional signal intensity data. T₂* is influenced by susceptibility gradients at a microscopic level (reflecting the quantity and distribution of free water) where signal intensity is influenced by sequence parameters and hardware effects that can vary between scanners and even between scan sessions.⁸ Furthermore, the SI values must be normalized to a standard in the field of view, in this case posterior femoral cortical bone. With the SI method, finding an easy to identify and repeatable structure in the field of view, which is not potentially influenced by treatment effects, is difficult. Conversely, T₂* is an inherent tissue property so a normalization standard is unnecessary, eliminating any associated confounding effects. Thus, T₂*, with additional validation, may provide a more consistent assessment of the healing ligament across scanners of the same magnet strength than examination of signal intensity and could make the T₂* predictions for scans performed across scanning intervals, studies, and institutions readily comparable.

Our findings are supported by previous research showing that volume normalized to a different relaxation time parameter (T₂) correlates to ACL reconstruction graft structural properties.²³ However, this previous study was limited by a non-significant correlation between structural properties and T₂ values as an independent quantifiable variable. The lack of correlation found between T₂ and structural properties was likely the result of relatively long echo times (>10ms) used to collect T₂ and may have been too long for gathering the short relaxation times of graft tissue.²⁴ Building on these earlier findings, the current study found that by utilizing T₂*, we were able to better capture the shorter relaxation times that are inherent to organized structures such as ligaments and tendons.¹⁰ Furthermore, we were able to identify a significant individual contribution of T₂* in terms of ligament sub-volume defined by ranges of T₂* values (see Supplement 2). Additionally, a study using Ultra-short echo time (UTE) imaging found a correlation between level of collagen organization in the meniscus and T₂*. Short T₂* was correlated with more densely packed collagen fibers and longer T₂* values were associated with less densely packed collagen and meniscal damage.¹⁰ While our study did not utilize UTE imaging, tissues with shorter T₂* values were linked to higher ligament structural properties. This suggests that T₂* can serve as a proxy for tissue organization and remodeling can be identified with the MR methodology presented herein.

This study was limited by the use of only two echo times which adds uncertainty to the determination of relaxation time.⁸ A two-point determination can overestimate relaxation time and is more sensitive to SI variability.^8,25 However, the correlation analysis between the median T₂* and structural properties incorporates the inter-specimen variability in T₂*. Additionally, this approach served to limit scan time (19 min), an important factor when developing a technique for clinical use. The standard errors for the structural properties predictions are comparable to the previous methods using signal intensity,⁷ indicating an effective determination of T₂* (Fig. 2). A two point determination of T₂ has also been used in cartilage imaging^15,26 to limit scan time and maximize resolution when considering geometry was paramount, similar to the goal of our study. Further refinement of our imaging protocol could be achieved by collecting additional shorter echo times, which would allow for improved certainty in the T₂* estimation using a nonlinear least-squares fit of voxel intensity versus echo time. Including additional shorter echo times could minimize the population of tissue with a zero ms relaxation time and make the T₂* estimation less sensitive to signal noise. Also, as with any MR study, the tradeoffs between resolution, scan time and image acquisition parameters, such as echo time, must be considered.²⁶ Due to the relatively small size of the healing ACL, the imaging protocol was optimized to reconstruct the whole volume of the ligament, opposed to a single mid-substance slice,^3,23 and to minimize the influence of partial volume effects. To minimize these concerns we employed high resolution 3-D MR images with minimal voxel size (0.85×0.27×0.27 mm) and no slice gap (0 mm). Also, this study did not include histological confirmation of tissue organization level within the healing ligament. However, a previous meniscal study reported a link between T₂* and collagen organization.¹⁰ Lastly, the MR images collected in this study were collected postmortem eliminating problems with motion artifact. Future work will incorporate histological confirmation of tissue organization and evaluate the efficacy of using a T₂* imaging method to predict structural properties in vivo for longitudinal studies.

The results reported here provide a critical step toward the improvement of a noninvasive method for predicting the structural properties of a healing ligament for in vivo use. Furthermore, the use of relaxation time instead of signal intensity specifically makes this approach independent of image acquisition parameters and is the first step to making this approach comparable across institutions. This will be advantageous for reducing the number of animals used in pre-clinical studies, and provide a noninvasive outcome measure of ligament healing in multicenter clinical trials.