Femoral Cartilage Ultrasound Echo-Intensity Is a Valid Measure Of Cartilage Composition

Abstract

This study aimed to create a conversion equation that accurately predicts cartilage magnetic resonance imaging (MRI) T2 relaxation times using ultrasound echo-intensity and common participant demographics. We recruited 15 participants with a primary anterior cruciate ligament reconstruction between the ages of 18–35 years at 1–5 years after surgery. A single investigator completed a transverse suprapatellar scan with the ACLR limb in max knee flexion to image the femoral trochlea cartilage. A single reader manually segmented the femoral cartilage cross-sectional area to assess the echo-intensity (i.e., mean grey-scale pixel value). At a separate visit, a T2 mapping sequence with the MRI beam set to an oblique angle was used to image the femoral trochlea cartilage. A single reader manually segmented the cartilage cross-sectional area on a single MRI slice to assess the T2 relaxation time. A stepwise, multiple linear regression was used to predict T2 relaxation time from cartilage echo-intensity and common demographic variables. We created a conversion equation using the regression betas and then used an ICC and Bland-Altman plot to assess agreement between the estimated and true T2 relaxation time. Cartilage ultrasound echo-intensity and age significantly predicted T2 relaxation time (F=7.33, p=0.008, R²=0.55). When using the new conversion equation to estimate T2 relaxation time from cartilage echo-intensity and age, there was strong agreement between the estimated and true T2 relaxation time (ICC_2,k=0.84). This study provides promising preliminary data that cartilage echo-intensity combined with age can be used as a clinically accessible tool for evaluating cartilage composition.

INTRODUCTION

Patients following anterior cruciate ligament (ACL) injury and reconstruction (ACLR) are at an increased risk for developing knee osteoarthritis (OA).¹ Specifically, undergoing an ACLR leads to a 6 times increased risk of OA within 11 years and 6 times increased lifetime risk of undergoing a knee arthroplasty.^{2, 3} Articular cartilage alterations are a hallmark sign of OA and compositional changes to the cartilage often precede overt changes in cartilage thickness during the early phases of OA development.^{4, 5} Compositional magnetic resonance (MR) sequences (e.g., T2 mapping and T1rho) are the current gold standard for the in vivo assessment of cartilage composition.^4–6 T2 relaxation times are theorized to provide an assessment of cartilage content and organization, while T1rho relaxation times are theorized to assess proteoglycan density.⁷ Previous studies have demonstrated that cartilage composition alterations exist within months after ACL injury and ACLR.^8–10 Additionally, early measures of cartilage composition are related to future declines in patient reported and functional outcomes following ACLR.^11–14 Despite the potential importance of MR-derived outcomes of cartilage composition as an early prognostic biomarker of OA, there are limitations in cost, accessibility, and technical expertise that prevent the widespread use of advanced compositional imaging techniques outside of well-funded research environments.

Point-of-care ultrasound is a more cost-effective and clinically accessible imaging modality when compared to MR imaging and has been validated as an alternative for assessing morphology in femoral cartilage thickness.¹⁵ Previous ultrasound studies have demonstrated greater femoral cartilage thickness in participants with an ACLR when compared to their contralateral limb and a limb from uninjured controls.¹⁶ In addition to assessing thickness, cartilage ultrasound assessments provide a measure of the image echo-intensity (i.e., brightness). Ultrasound echo-intensity is frequently used as a surrogate for muscle composition in prior studies,¹⁷ but has only recently been used as a potential indicator of cartilage health.^18–20 For example, the presence of arthroscopically-assessed cartilage damage in patients undergoing ACLR was associated with lower ultrasound-assessed cartilage echo-intensity.²⁰ Additionally, pre-operative femoral cartilage echo-intensity is associated with early OA symptoms in participants 1 year post-ACLR.¹⁹ While these studies provide preliminary importance of cartilage echo-intensity as a measure of early cartilage alterations, femoral cartilage echo-intensity has never been validated to MR imaging measures of cartilage composition. Additionally, there is evidence that ultrasound echo-intensity and MRI cartilage composition may be influenced by participant demographics,^{17, 21} which may need to be accounted for when validating ultrasound echo-intensity to MRI cartilage composition.

Therefore, the purpose of this study was to validate cartilage ultrasound echo-intensity against a compositional MR assessment (i.e., T2 relaxation times). Since ultrasound echo-intensity and cartilage composition may be influenced by patient-specific demographics,^{17, 21} our models included both cartilage ultrasound echo-intensity, as well as common demographic variables to create a conversion equation to estimate T2 relaxation times. We then determined the agreement between the estimated T2 relaxation times and true, MR-assessed T2 relaxation times using this conversion equation. We hypothesized that cartilage ultrasound echo-intensity would be significantly associated with MR-assessed T2 relaxation times. These results would be important as it would advance the evidence in support of ultrasound as a valid alternative to MRI for assessing cartilage composition in patients after ACLR.

METHODS

Study Design

We recruited a sample of 15 participants in this initial preliminary cohort to validate this novel application of ultrasound echo intensity for compositional analysis. This sample size is consistent with prior foundational imaging validity and feasibility studies in knee cartilage that used samples of 8–10 participants.^{15, 22} Participants in this cross-sectional study completed two data collection sessions that were separated by one week: 1) femoral trochlea cartilage ultrasound assessment, 2) T2 Mapping MR imaging sequence of the femoral trochlea. The time of testing the two data collection sessions were scheduled within 2 hours on the respective days to control for any diurnal effects on the cartilage.^{23, 24}

Participants

We recruited 15 participants between 1–5 years after a primary, unilateral ACLR. Participants were between 18–35 years old without any additional lower extremity surgery or diagnosis of any form of arthritis. We selected this age range because it is recommended for post-ACLR OA prevention clinical trials.²⁵ We excluded participants if they had a cardiovascular, neurological, or other medical condition that would prevent them from participating in our ongoing longitudinal study that involved strength and movement biomechanics assessments. All participants signed a written informed consent prior to study enrollment. Prior to participant enrollment, the local University’s Institutional Review Board approved all data collection procedures.

Acquiring the Femoral Cartilage Ultrasound Images

The ultrasound assessment was performed in a research laboratory by a single investigator with 8 years of cartilage ultrasound research experience using a GE LOGIQ P9 R3 ultrasound system and L3–12-RS wideband linear array probe (GE Healthcare, Chicago, IL).^{16, 18–20, 26–31} All ultrasound settings were kept consistent between participants. We used a depth of 4 cm, a gain of 50, with an image resolution of 212 pixels per 1 cm, and a probe with a 5.5 cm width. The participants were positioned supine on a plinth for at least 30 minutes prior to the cartilage ultrasound assessment.²⁰ Three images of the femoral trochlear cartilage were recorded on the ACLR limb with the participants supine and their ACLR limb in maximal knee flexion (137±3.8°) using a transverse suprapatellar ultrasound scan. The ultrasound probe was placed in line with the apex of the medial and lateral femoral condyles then rotated inferiorly and superiorly to identify the probe orientation where the cartilage borders were most distinct (Figure 1A).^{19, 20} The ultrasound probe was removed and repositioned prior to acquiring all three images.

Figure 1. Femoral Cartilage Ultrasound Image Acquisition.

A) Participant and ultrasound probe positioning for the cartilage ultrasound assessment. B) Example image and segmentation (white region) of the femoral trochlear cartilage acquired with a transverse suprapatellar ultrasound scan.

Segmenting the Femoral Cartilage Ultrasound Images and Calculating Echo-Intensity

A single, trained reader manually segmented the femoral cartilage cross-sectional area of each image using the publicly available ImageJ software (https://imagej.nih.gov/; Figure 1B).³² The main ultrasound outcome used in this study was echo-intensity, which is the mean grey-scale pixel value ranging from 0 (i.e., black) to 255 (i.e., white) within the segmented cartilage area. We have previously demonstrated the test-retest reliability of cartilage ultrasound echo-intensity across multiple images within a single visit (ICC_2,k=0.93–0.94), as well as between two visits separated by at least one week (ICC_2,k=0.83–0.95) in a young active population.¹⁸ To highlight the precision of assessing cartilage ultrasound echo-intensity, we also have previously established the test-retest standard error of the measurement (SEM=1.00–1.41 au) and the minimal detectable change (MDC = 2.11–3.29 au).¹⁸

Acquiring the T2 Mapping MR Images

Compositional knee MR images were obtained using a T2 Mapping sequence that were acquired in a university research imaging facility with an 8-channel phased array knee coil (GE 3 Tesla Excite MR Imaging scanner). The participants were positioned supine in the MR scanner for at least 30 minutes prior to the acquisition of the compositional MR sequence.³³ MRI imaging was done in a supine position with slight knee flexion. Oblique slices (3 mm slice thickness) were oriented perpendicular to the femoral condyles to provide a similar view as the ultrasound assessment to assess the femoral trochlear cartilage (Figure 2). We used a T2 Mapping sequence with 8 different spin-lock durations (fast spin echo, TR=1500ms, TE =9.5, 18.9, 28.7, 37.8, 47.3, 56.7, 66.2, 75.6 ms) with a 256×256 matrix yielding a voxel size of 0.625 mm × 0.625 mm × 3 mm. We created voxel by voxel T2 relaxation time images from the 8-image sequence using a custom MatLab program (Matlab R2020, MathWorks, Matick, MA) using the following equation: S(TSL) = S0 exp(−TSL/T2). TSL is the duration of the spin-lock time, S0 is the signal intensity when TSL equals zero, S is the signal intensity, and T2 is the T2 relaxation time in the rotating frame.

Figure 2. Acquiring Oblique Angle Magnetic Resonance Images to View the Femoral Trochlear Cartilage.

A) Example image of the orientation of the oblique MR slices through the femoral trochlea. We utilized the middle MR slice, as this represents the slice that passes most perpendicularly through the trochlear cartilage. B) Example image of the femoral trochlear cartilage using the oblique MR slice with the corresponding oblique slice orientation in the bottom right.

Segmenting the Femoral Cartilage on the T2 Mapping MR Images

A single, different trained reader, blinded to the ultrasound results, manually segmented the femoral cartilage cross-sectional area on a single slice of the T2 relaxation time image using ITK-Snap (www.itksnap.org; Figure 2B).³⁴ As seen in Figure 2, the MRI T2 image was perpendicular with the trochlear cartilage. The reader selected the middle MR slice that was specifically oriented to be perpendicular to the trochlear cartilage. Additionally, the shape of the femoral trochlear cartilage in this middle MR slice most resembled the shape of the femoral trochlear cartilage acquired with ultrasound (Figure 2B). The main MR outcome was T2 relaxation time, which was the mean intensity of cartilage within the segmented region of the T2 Mapping MR slice.

Demographic Variables using in the Analysis

Demographic variables including age, body mass, height, sex, and time since ACL reconstruction were collected through participant surveys and researcher assessments. Age was calculated between date of testing and self-reported participant birthday. Height was measured to the nearest 0.1 cm using a stadiometer. Body mass was measured to the nearest 0.1 kg using a calibrated scale. Sex was self-reported. Time since ACLR was calculated between date of testing and self-reported date of ACLR. All demographic variables were collected in the same manner for each participant during the initial study visit. We selected these variables as these are likely going to be recorded in any study using ultrasound or MRI to assess cartilage composition, as well as because each of these may also be related to cartilage composition.^{17, 21} The inclusion of these variables in the model is hypothesized to provide a more patient-population specific conversion equation to accurately estimate T2 relaxation times in participants post-ACLR.

Statistical Analysis

Prior to our primary regression analysis, bivariate Pearson correlations were calculated between each predictor variables (i.e., echo-intensity and demographics) and the outcome T2 relaxation times. A forward stepwise multiple linear regression was used to predict the T2 relaxation time from cartilage echo-intensity and common demographic variables (i.e., age, mass, height, sex, months post ACLR). Variables were added to the model one at a time based on the significance of the F-value (entry p<0.05), with variables removed if they became non-significant (removal p>0.10). We chose a stepwise regression approach to identify the most parsimonious model predicting T2 relaxation times from the set of candidate variables. This was deemed advantageous for an initial analysis given the limited sample size. Using the regression betas for the included outcomes in the regression model with the highest R², we calculated an estimated T2 relaxation time variable. We then determined the agreement between the estimated and true T2 relaxation times using an absolute agreement intraclass correlation coefficient (ICC_2,k) with 95% confidence intervals (CI) and a Bland-Altman plot. The ICCs were classified as weak (<0.5), moderate (0.5–0.69), or strong (≥0.7).³⁵ The Bland-Altman plots graph the difference between the estimated and true T2 relaxation times on the y-axis against the average of the estimated and true T2 relaxation times.³⁶ An upper and lower 95% limit of agreement was determined as 1.96 times the standard deviation of the mean differences. We defined acceptable agreement as no more than 5% of all data points located outside of the 95% limits of agreement,³⁶ as well as calculating the mean difference between the estimated and true T2 relaxation times.

RESULTS

Of the 15 included participants, the majority were female (n=8), with an average height of 173±9 cm (range: 158–188.5), mass of 71.1±15.1 kg (range: 50.3–100.7), age of 25.8±5.7 years (range: 18–35), and were 29±12 months post-ACLR (range: 12–53). The participants had average T2 relaxation times of 51.1±2.7ms and cartilage echo-intensity of 38.1±5.1 arbitrary units. In the univariate correlational analysis, T2 relaxation times was significantly associated with cartilage echo-intensity (r=0.52, p=0.048) and age (r=0.59, p=0.022), but not height (r=0.07, p=0.805), mass (r=−0.067, p=0.811), sex(r=−0.133, p=0.638), or months post ACLR (r=0.21, p=0.455). Similarly, the forward stepwise multiple linear regression indicated that cartilage echo-intensity and age combined to significantly predict T2 relaxation times (F=7.33, p=0.008, R²=0.55). Table 1 highlights the unstandardized betas used in the equation to convert cartilage echo-intensity and age to an estimated T2 relaxation time.

Table 1.

Regression Beta Coefficients used in T2 Relaxation Time Conversion Equation

Variable	Unstandardized Beta	Standardized Beta	t	p
(Constant)	35.106
Age (years)	0.254	0.53	2.74	0.018
Echo-Intensity (AU)	0.247	0.46	2.34	0.037

Dependent Variable = MRI T2 relaxation time

Estimated T2 Relaxation Time = 35.106 + (0.254*Age) + (0.247*Cartilage Echo-Intensity)

AU = arbitrary units. Pixel echo-intensity ranges from 0 (black) – 255 (white)

$$\underset{¯}{\text{Estimated}\text{T2}\text{Relaxation}\text{Time}}=35.106+(0.254*\text{Age})+(0.247*\text{Cartilage}\text{Echo-Intensity})$$ Equation:

The ICC analysis indicates strong agreement (ICC_2,k=0.84, 95% CI: 0.51–0.95) between the estimated T2 relaxation time calculated using our conversion equation that includes cartilage echo-intensity and age, and the true T2 relaxation time values. The Bland-Altman plot indicates there is acceptable agreement and no mean difference or bias between the estimated and true T2 relaxation times (Figure 3).

Figure 3. Bland-Altman Plot comparing the Estimated and True T2 Relaxation Times.

The Bland-Altman plot depicts the difference between the true and estimated T2 relaxation times (y-axis) against the mean of the true and estimated T2 relaxation times (x-axis) for each participant. The Bland-Altman plot highlights that the mean of the difference between the true and estimated T2 relaxation times was 0.00ms. The 95% limits of agreement were −3.81 ms and 3.60 ms and no participants fell outside these limits of agreement.

DISCUSSION

Our results demonstrate that cartilage echo-intensity on ultrasound images is associated with a gold standard, in vivo metric of cartilage composition. We developed a novel conversion equation that includes cartilage ultrasound echo-intensity and age to estimate T2 relaxation times in patients 1–5 years after ACLR. The estimated T2 relaxation time values that were calculated using the conversion equation were in strong agreement and had a minimal mean difference when compared to the true T2 relaxation times assessed via MR. This study provides promising preliminary data that cartilage echo-intensity can be used as a clinically accessible tool for evaluating cartilage composition.

Cartilage alterations in the femoral trochlea are prevalent within the first year following ACLR and are prognostic of future reports of increased pain, worse symptoms, and decreased quality of life.^37–39 Point-of-care ultrasound is a valid and reliable imaging modality that can be used to quantitatively assess femoral cartilage morphology (e.g., thickness).¹⁵ Recent reports have used ultrasound to highlight femoral cartilage thickness alterations in participants following ACLR,^{16, 19, 20, 27, 40–43} but these studies have mostly been limited to an assessment of cartilage morphology. In addition to cartilage morphology, the segmentation of cartilage ultrasound images also provides a quantification of the image echo-intensity (i.e., image brightness) as an indirect biomarker of cartilage composition. Previous ultrasound studies provided initial justification that the signal intensity of ultrasound may be related to cartilage composition,^{44, 45} but this is the first study to validate an in vivo assessment of cartilage ultrasound echo-intensity against a gold standard MR assessment of cartilage composition. This adds to recent studies that highlight the association of cartilage ultrasound echo-intensity with arthroscopic cartilage damage and early OA symptoms in participants following ACL injury and reconstruction.^{19, 20} Similar to the findings of this study, cartilage ultrasound echo-intensity when assessed pre-operatively was associated with an intraoperative, arthroscopic assessment of cartilage damage in patients undergoing ACLR.²⁰ Additionally, participants that meet a classification criteria for early OA symptoms at 1 year post-ACLR have altered femoral cartilage ultrasound echo-intensity pre-operatively, when compared to patients that do not report early OA symptoms.¹⁹ Collectively, these studies highlight the construct validity of using cartilage ultrasound echo-intensity as a clinically accessible measure of MR T2 relaxation times (i.e., collagen content and organization), as well as the prognostic capability of an early assessment of cartilage ultrasound echo-intensity in patients at risk for OA.

Due to knee joint anatomy, a traditional transverse, suprapatellar ultrasound assessment of femoral cartilage is limited to assessing the femoral trochlear cartilage. The femoral trochlea is extremely relevant to early OA after ACLR, as this location exhibits the earliest and most rapid decline in cartilage health after ACLR.^{37, 38, 46–48} Therefore, despite the femoral cartilage ultrasound assessment being limited to trochlea, this is a clinically significant region of the knee that is an optimal location for monitoring early OA. However, traditional cartilage regions assessed with MR are typically focused on the tibiofemoral joint with MR slices in the sagittal or frontal plane. Due to this, the MR slices in the femoral trochlea are not perpendicularly aligned with the femoral trochlea cartilage, resulting in a different view of the cartilage between MR and ultrasound. Previous ultrasound validation studies attempt to associate the ultrasound-assessed cartilage to the traditional tibiofemoral MR regions of interest.²² However, cartilage morphology and composition differ vastly throughout the different regions of the femur,⁴⁹ which may bias the results of a validation study because the ultrasound and MR assessments are not assessing the same cartilage. Therefore, we used an obliquely oriented MR slice through the femoral trochlea to align with the cartilage assessed via the traditional femoral cartilage ultrasound assessment.

While this study provided further evidence that cartilage ultrasound echo-intensity is a clinically accessible technique to assess femoral trochlea cartilage composition, there are some limitations to discuss. The transverse suprapatellar ultrasound image acquired in this study is limited to the assessment of the femoral trochlea and unable to image the weight-bearing tibiofemoral cartilage due to the bony anatomy of the knee. However, the femoral trochlea is often the first and most severely affected region of femoral cartilage in participants after ACLR. ^{37, 38, 46–48} We used an obliquely angled MR slice in order to try to replicate a similar location of cartilage assessed between the MR and ultrasound assessments. However, a key limitation is the lack of confirmation that identical anatomical regions were assessed with MRI and ultrasound due to differences in knee positioning. While oblique MRI planes aimed to align with the transverse ultrasound view, inherent challenges exist in confirming the localization. More advanced image registration and spatial normalization techniques could be applied in future studies to allow for definitive cartilage region matching between the modalities. Since age is one of the main predictors in our results, this conversion equation should only be applied to participants within the age range of the included participants (i.e., age range: 18–35 years). The other participant demographics included in our regression analysis (i.e., mass, height, sex, and time since ACLR) were not significant predictors in our final model estimating T2 relaxation times from echo-intensity. This may be due to our small sample size lacking power to detect weaker associations with these factors. The sample size in this validation study was relatively small (n=15) and future studies may be needed to confirm the accuracy of this cartilage composition conversion equation in independent samples. Our MRI protocol focused on a T2 Mapping sequence so that we could validate ultrasound echo-intensity to a measure of cartilage composition. Since we did not include any morphological sequences, we are not able to estimate cartilage thickness in this region. However, prior studies have provided initial evidence of the validation of cartilage thickness assessment using ultrasound.^{15, 22} This preliminary study did not include comparisons to other modalities like histology or biochemical analysis that can also characterize cartilage composition. Future studies could validate ultrasound echo intensity against these standards along with MRI.

In conclusion, this study indicates that cartilage echo-intensity coupled with age can be used as a clinically accessible tool for evaluating femoral cartilage composition. The conversion equation that includes cartilage ultrasound echo-intensity and age can be used in future studies that want to calculate an indirect measure of cartilage T2 relaxation times. This will be important for future studies that use ultrasound: to monitor joint health in patients at-risk for knee OA development, as an inclusion criteria in clinical trials, or to quantify changes in cartilage health after OA prevention interventions.