Intravoxel Incoherent Motion (IVIM) Detects Femoral Head Ischemia in a Piglet Model of Legg-Calvé-Perthes Disease

Abstract

There is a clinical need for alternatives to gadolinium contrast-enhanced magnetic resonance imaging (MRI) to facilitate early detection and assessment of femoral head ischemia in pediatric patients with Legg-Calvé-Perthes disease (LCPD), a juvenile form of idiopathic osteonecrosis of the femoral head. The purpose of this study was to determine if intravoxel incoherent motion (IVIM), a non-contrast-enhanced MRI method to simultaneously measure tissue perfusion and diffusion, can detect femoral head ischemia using a piglet model of LCPD. Twelve six-week-old piglets underwent unilateral hip surgery to induce complete femoral head ischemia. The unoperated, contralateral femoral head served as a perfused control. The bilateral hips of the piglets were imaged in vivo at 3T MRI using IVIM and contrast-enhanced MRI one week after surgery. Median apparent diffusion coefficient (ADC) and IVIM parameters (diffusion coefficient: D_s; perfusion coefficient: D_f; perfusion fraction: f; and perfusion flux: f*D_f) were compared between regions of interest comprising the epiphyseal bone marrow of the ischemic and control femoral heads. Contrast-enhanced MRI confirmed complete femoral head ischemia in 11/12 piglets. IVIM perfusion fraction (f) and flux (f*Df) were significantly decreased in the ischemic vs. control femoral heads: on average, f decreased 47±27% (Δf=−0.055±0.034; p=0.0003) and f*Df decreased 50±27% (Δf*Df=−0.59±0.49×10⁻³ mm²/s; p=0.0026). In contrast, IVIM diffusion coefficient (D_s) and ADC were significantly increased in the ischemic vs. control femoral heads: on average, D_s increased 78±21% (ΔD_s=0.60±0.14×10⁻³ mm²/s; p<0.0001) and ADC increased 60±36% (ΔADC=0.50±0.23×10⁻³ mm²/s; p<0.0001). In conclusion, IVIM is sensitive in detecting bone marrow ischemia in a piglet model of LCPD.

GRAPHICAL ABSTRACT

IVIM is a non-contrast-enhanced MRI technique sensitive to tissue perfusion and a potential alternative to gadolinium contrast-enhanced MRI (CE-MRI). Using a piglet model of LCPD, a pediatric hip disorder caused by interruption of blood supply to the femoral head, we found that IVIM-measured perfusion fraction (f) was decreased in ischemic vs. perfused femoral heads. Our findings support that IVIM may be an alternative to CE-MRI to assess bone marrow perfusion in patients with LCPD and other ischemic bone disorders.

INTRODUCTION:

Legg-Calvé-Perthes disease (LCPD) is a childhood hip disorder caused by interruption of blood supply to the developing femoral head, which can lead to osteonecrosis, deformation of the femoral head, and the onset of hip osteoarthritis at a young age.¹ The ischemic injury results in necrosis of bone marrow and bone cells in the avascular stage of the disease, halting the growth of the secondary ossification center (SOC; i.e., the epiphyseal bone and bone marrow) of the femoral head. This is then followed by a repair stage that includes spontaneous revascularization, resorption of the necrotic bone, and, eventually, formation of new bone in an attempt to restore the femoral head. The goal of clinical management of LCPD is to preserve the congruency of the hip joint through preventing or minimizing femoral head deformity. Surgical interventions aim to increase containment of the femoral head in the acetabulum and may also impact the reparative biology of LCPD, for example by increasing the speed of the repair.^2,3 A key factor in clinical decision-making is the perfusion status of the femoral head, which informs the extent and stage of the disease and plays a significant role in determining whether surgery is recommended.^4–6

Subtraction contrast-enhanced magnetic resonance imaging (CE-MRI) utilizing intravenous injection of gadolinium contrast material is the primary method used clinically to assess the extent of ischemia and/or reperfusion in early-stage LCPD.⁷ However, the method is used sparingly due to concerns about the unknown long-term effects of gadolinium contrast agent use in pediatric patients as well as the need for sedation of children to get high-quality pre- and post-contrast subtractions.^7,8 Thus, there is a clinical need for an alternative, non-contrast-enhanced approach to measure femoral head perfusion in patients with LCPD. A non-contrast-enhanced method would also allow MRI to be used more frequently, facilitating serial exams, which may not only inform treatment decisions but also lead to better understanding of the complex pathogenesis of LCPD.^1,9 Quantitative MRI techniques, including apparent diffusion coefficient (ADC) mapping using diffusion-weighted imaging (DWI) and T2 and T1ρ relaxation time mapping, have been proposed as alternative or complementary approaches to CE-MRI to assess early-stage LCPD.^10–18 For example, ADC has been shown to increase in the SOC in response to ischemic injury to the bone marrow of the femoral head.^10–12 While these methods may be useful to assess the severity and extent of the injury and/or repair to the femoral head, they do not provide information on whether or not the bone marrow is perfused.

Intravoxel incoherent motion (IVIM) is a non-contrast-enhanced MRI technique with the potential to measure femoral head perfusion. IVIM uses a multi-b-value DWI acquisition and a two-compartment model to simultaneously quantify the tissue diffusion coefficient (D_s), perfusion coefficient (D_f), perfusion fraction (f), and perfusion flux (f*D_f).¹⁹ IVIM has been utilized to evaluate the function of multiple highly-perfused organs and tissues such as the brain, liver, pancreas, kidney, and prostate,^20,21 and it can measure a decrease in brain perfusion fraction (f) in patients suffering from stroke.²² However, low-perfused tissues (such as bone marrow) present a challenge for IVIM due to difficulties in reliably fitting for low f values of perfusion fraction;²³ for example, prior studies have reported low reliability of IVIM-measured perfusion in the bone marrow of lumbar vertebrae.^24,25 IVIM also has not been validated for imaging of bone marrow ischemia; while a prior study of children being treated for developmental dysplasia of the hip measured reduced perfusion in affected femoral heads using IVIM,²⁶ these observations were not validated by comparing to a gold-standard technique such as CE-MRI. Thus, there is a need to further validate and improve upon the IVIM technique to image bone marrow perfusion and ischemia.

The purpose of the current study was to determine if IVIM can detect femoral head ischemia using a piglet model of LCPD with CE-MRI as the gold standard. In the piglet model, one hip joint underwent surgical induction of complete ischemia to the bony epiphysis of the femoral head, while the contralateral femoral head served as the unoperated, perfused control. The piglet model thus enabled a controlled experiment to determine whether femoral head ischemia and perfusion are detectable using IVIM. We hypothesized that IVIM-measured perfusion fraction (f) and flux (f*D_f) would be decreased in the ischemic vs. control femoral heads one week after surgery. We also compared changes in the IVIM diffusion coefficient (D_s) and ADC to assess their relative sensitivity to detect ischemic injury to the SOC.¹¹ This study is an important step to validate the sensitivity of IVIM in detecting bone marrow ischemia and perfusion and to inform its potential clinical translation to assess patients with LCPD and other ischemic bone and joint disorders.

METHODS:

Animal Model:

This study was approved by the University of Minnesota’s Institutional Animal Care and Use Committee. Twelve Yorkshire piglets (8 male, 4 female) were obtained from a commercial provider (Manthei Hog Farm, LLC; Elk River, MN) and housed in pairs throughout the study in the University of Minnesota’s Research Animal Resources facilities. The sample size was chosen to limit the use of animals while being sufficient to detect a large effect size of 1.0 with 80% statistical power and a 5% significance level. Piglets underwent surgery at six weeks of age (mean weight = 10.2±1.8 kg and range = 8.1 to 14.5 kg) to induce complete femoral head ischemia by placing a ligature around the femoral neck and transecting the ligamentum teres; this procedure results in radiological and histological changes resembling those of LCPD.^27,28 No surgery was performed on the contralateral femoral head, which served as a paired, perfused control. Bilateral hips of the piglets were imaged in vivo at 3T MRI one week following surgery. Piglets were excluded from the study if they did not have complete (100%) femoral head ischemia at the time of the in vivo MRI exam, as confirmed by lack of signal enhancement on subtracted CE-MRI, which was acquired after the IVIM acquisition. For the surgical and MRI procedures, piglets were sedated using intramuscular administration of either a combination of midazolam (10 mg/kg) and buprenorphine (10 μg/kg) or a combination of telazol (4.0 mg/kg) and xylazine (2.0 mg/kg). Anesthesia was induced by intravenous administration of either propofol (2.0-6.0 mg/kg) or both ketamine (5.0 mg/kg) and propofol (2.0-6.0 mg/kg). Anesthesia was maintained by insufflation of isoflurane (1.0-5.0%) vaporized in oxygen. Postoperative pain control was provided by oral administration of carprofen (2.0-3.0 mg/kg) once or twice daily for three days.

In Vivo 3T MRI:

In vivo imaging of the piglets was performed using a clinical 3T MRI scanner (MAGNETOM Prisma; Siemens Healthcare; Erlangen, Germany) and two vendor-provided four-channel flex receiver arrays. First, multi-b-value images for IVIM were acquired using a RESOLVE (readout segmentation of long variable echo trains) DWI sequence with the following parameters: FOV = 208×208 mm²; sampling matrix = 188×188; spatial resolution = 1.1×1.1 mm²; slices = 15; slice thickness/gap = 2.0/0.5 mm; TR/TE₁/TE₂ = 2500/68/122 ms; echo spacing = 0.5 ms; EPI factor = 94; bandwidth = 578 Hz/px; readout segments = 5; GRAPPA = 2; b-values = 0, 20, 30, 40, 50, 60, 70, 80, 90, 100, 300, and 500 s/mm²; 3 diffusion directions; monopolar diffusion gradients; fat saturation; and scan time = 10 or 21 minutes (6/11 data sets had additional b-value averages and thus a longer scan time). Subsequently, subtraction CE-MRI images were acquired to confirm surgical induction of complete ischemia by lack of contrast uptake in the bony epiphysis of the operated femoral head. For the subtraction, two identical pre- and post-contrast images were acquired using a 2D TSE sequence with parameters: FOV = 200×169 mm²; sampling matrix = 448×378; spatial resolution = 0.45×0.45; slices = 20; slice thickness/gap = 2.0/0.6 mm; TR/TE = 970/21 ms; flip angle = 150 degrees; averages = 2; turbo factor = 4; GRAPPA = 2; bandwidth = 260 Hz/px; and fat saturation. The post-contrast image was acquired following intravenous administration of 0.2 mmol/kg gadolinium contrast material (ProHance; Bracco Diagnostics; Monroe Township, NJ) and a 1-minute delay.

Image and Data Analysis:

Subtraction CE-MRI images of the femoral heads were examined, and only piglets with a complete absence of contrast uptake in the bony epiphysis of their operated femoral heads were included in the study.

Before the analysis of the DWI data, a denoising technique called Marchenko-Pastur Principal Component Analysis (MPPCA)²⁹ was applied to all acquired trace images to improve their signal-to-noise ratio. The apparent diffusion coefficient (ADC) was then estimated from a simple mono-exponential fit using all of the denoised b-value trace images (Equation 1):

$$S(b)=S(0)e^{-b·ADC}.$$ (1)

Classically, IVIM parameters are extracted using a two-compartment exponential model that fits the DWI signal (S) as a function of the b-value as follows¹⁹ (Equation 2):

$$S(b)=S(0)[f{e}^{-b(D_s+D_f)}+(1-f)e^{-b{D}_s}].$$ (2)

Since the baseline perfusion in the femoral head was anticipated to be low, we used a new analytical segmented (AS) fitting approach to fit the IVIM parameters, which is particularly useful for assessing low perfusion tissues.^30,31 In brief, the AS approach uses a segmented fitting approach based on the assumption that the perfusion coefficient (D_f) is at least ten times higher than the diffusion coefficient (D_s).¹⁹ Thus, the contribution of perfusion coefficient (D_f) at a sufficiently high b-value (b ≥ b_threshold) is negligible. Therefore, the diffusion coefficient (D_s) can be more reliably obtained by fitting images with b ≥ b_threshold using a simple mono-exponential function (Equation 3):

$$S(b)=S(0)e^{-b{D}_s},\cdot \mathrm{b}\ge {\mathrm{b}}_{\text{threshold}}.$$ (3)

For this study, we set ${\mathrm{b}}_{\text{threshold}}=100\mathrm{s}/{\mathrm{mm}}^2$, To measure the perfusion parameters, Equation 2 can be rearranged, as shown in Equation 4. Using the value of D_s from Equation 3, the left-hand side of Equation 4 can be calculated for every b-value. With these values, the mono-exponential growth function on the right-hand side of Equation 4 is then used to extract both f and D_f⋅ using all acquired b-values:

$$1-\frac{S(b)}{S(0)}{e}^{b{D}_s}=f\left(1-e^{-b{D}_f}\right).$$ (4)

ADC, D_s, D_f, and f quantitative maps were generated for each piglet using the AS fitting approach outlined above. Median ADC and IVIM parameters (diffusion coefficient: D_s; perfusion coefficient: D_f; perfusion fraction: f; and perfusion flux: f*D_f) were then calculated in two regions of interest (ROIs) comprising the SOCs (i.e., the bony epiphyses) of the (i) operated (ischemic) and (ii) contralateral-control (perfused) femoral heads. ROIs were defined using a single 2D central imaging slice from the b=0 diffusion-weighted and 2D TSE imaging series. A single slice was chosen to minimize through-plane partial volume averaging, and the ROIs were carefully drawn to avoid the epiphyseal cartilage overlying the SOC. The ROIs were drawn by an experienced observer (C.P.J.) who was blinded to the IVIM maps, and a second observer (E.O.B.) also independently drew the ROIs to test the reliability of the ROI definitions. The paired differences in the median ADC and IVIM values between the ischemic vs. control femoral heads were calculated and statistically compared using paired t-tests. The significance threshold was set to p<0.01 after Bonferroni correction for five quantitative measures (ADC, D_s, D_f, f, and f*D_f). We also assessed effect sizes (Cohen’s d) and percent changes for each measurement.

RESULTS:

Complete femoral head ischemia was confirmed in 11 of the 12 piglets as determined by complete absence of contrast uptake on subtracted CE-MRI. One piglet had only partial femoral head ischemia and thus was excluded from the data analysis.

IVIM measures of perfusion fraction (f) and flux (f*D_f) were significantly decreased in the SOCs of the ischemic vs. control femoral heads. In the control femoral heads, the average perfusion fraction (f) was 0.109±0.043 and the average perfusion flux (f*D_f) was 1.10±0.41 ×10⁻³ mm²/s (Table 1). In the ischemic femoral heads, perfusion fraction (f) and flux (f*D_f ) were decreased on average by 0.055±0.034 (p=0.0003) and 0.59±0.49 ×10⁻³ mm²/s (p=0.0026), respectively, which correspond to effect sizes of d=1.6 and 1.2 and percent changes of −47±27% and −50±27%, respectively (Table 1 and Figure 1). Perfusion measurements for the individual piglets are shown in Table 2 and Figure 2: Perfusion fraction (f) decreased in the ischemic vs. control femoral head in 9/11 piglets (2/11 cases had negligible difference), and perfusion flux (f*D_f) decreased in 10/11 piglets (1/11 cases showed an increase). The IVIM perfusion coefficient (D_f) did not show a significant difference between the ischemic and control femoral heads.

Table 1:

Paired differences in the ADC and IVIM parameter measurements in the SOCs of the ischemic and control femoral heads across n=11 piglets.

		Control	Ischemic	Paired Difference (Δ)	95% CI	t Value	p Value	Effect Size (Cohen’s d)	Percent Change (%)
Diffusion	ADC (×10−3 mm2/s)	0.93 ± 0.20	1.43 ± 0.14	0.50 ± 0.23	[0.34, 0.65]	7.2	<0.0001 *	2.2	60 ± 36
	Ds (×10−3 mm2/s)	0.78 ± 0.10	1.38 ± 0.15	0.60 ± 0.14	[0.50, 0.69]	14.6	<0.0001 *	4.4	78 ± 21
Perfusion	f	0.109 ± 0.043	0.055 ± 0.036	−0.055 ± 0.034	[−0.077, −0.032]	5.4	0.0003 *	1.6	−47 ± 27
	Df (×10−3 mm2/s)	17.2 ± 9.8	19.9 ± 8.4	2.7 ± 9.3	[−3.6, 9.0]	1.0	0.36	0.3	34 ± 66
	f*Df (×10−3 mm2/s)	1.10 ± 0.42	0.52 ± 0.38	−0.59 ± 0.49	[−0.94, −0.23]	4.0	0.0026 *	1.2	−50 ± 27

Values are reported as mean ± standard deviation across the piglets’ median SOC values.

^*Statistically significant (p<0.01)

ADC = apparent diffusion coefficient; D_s = diffusion coefficient; f = perfusion fraction; D_f = perfusion coefficient; and f*D_f = perfusion flux

Figure 1:

Average percent change in ADC and IVIM parameters within the SOC of the ischemic vs. control femoral heads (n=11 pairs). ADC and IVIM diffusion coefficient (D_s) were significantly increased, while IVIM perfusion fraction (f) and perfusion flux (f*D_f) were significantly decreased. Error bars indicate the standard deviation. * p<0.01

Table 2:

Differences (Δ) and percent changes in the quantitative values of ADC and IVIM parameters in the SOCs of the ischemic vs. control femoral heads for each animal.

Piglet #	ΔADC (×10−3 mm2/s)	ΔDs (×10−3 mm2/s)	Δf	ΔDf (×10−3 mm2/s)	Δf*Df (×10−3 mm2/s)
1	+0.79 (+107%)	+0.79 (+107%)	−0.033 (−50%)	−14.1 (−34%)	−0.24 (−39%)
2	+0.69 (+98%)	+0.65 (+94%)	−0.090 (−61%)	+5.23 (+44%)	−0.37 (−42%)
3	+0.83 (+94%)	+0.84 (+98%)	−0.042 (−55%)	+7.28 (+26%)	−0.76 (−59%)
4	+0.69 (+94%)	+0.70 (+104%)	−0.004 (−6%)	−6.40 (−33%)	−0.64 (−62%)
5	+0.11 (+10%)	+0.48 (+67%)	−0.068 (−49%)	+5.79 (+79%)	−0.24 (−37%)
6	+0.59 (+87%)	+0.59 (+91%)	−0.001 (−2%)	+1.08 (+7%)	−0.22 (−39%)
7	+0.341 (+39%)	+0.57 (+70%)	−0.093 (−73%)	+18.3 (+168%)	−0.65 (−58%)
8	+0.35 (+33%)	+0.46 (+49%)	−0.032 (−18%)	+15.8 (+124%)	+0.22 (+16%)
9	+0.38 (+33%)	+0.45 (+59%)	−0.072 (−54%)	−2.39 (−14%)	−1.35 (−76%)
10	+0.38 (+40%)	+0.47 (+59%)	−0.081 (−74%)	−2.63 (−15%)	−1.39 (−79%)
11	+0.27 (+22%)	+0.55 (+59%)	−0.087 (−75%)	+1.79 (+21%)	−0.82 (−78%)

Values are shown as the difference and, in parentheses, the percent change between the ischemic and control femoral heads: Δ (% change).

ADC = apparent diffusion coefficient; D_s = diffusion coefficient; f = perfusion fraction; D_f = perfusion coefficient; and f*D_f = perfusion flux

Figure 2:

Plots of the paired ADC and IVIM parameter measurements within the SOCs of the ischemic and control femoral heads for the 11 piglets. Each line shows the paired values for a specific piglet. Overall, ADC and IVIM diffusion coefficient (D_s) increased while IVIM perfusion fraction (f) and perfusion flux (f*D_f) decreased in the ischemic vs control femoral heads. Differences in the IVIM perfusion coefficient (D_f) were variable between animals.

In contrast to the changes in perfusion, measures of diffusion (IVIM diffusion coefficient D_s and ADC) were significantly increased in the SOCs of the ischemic vs. control femoral heads. On average, the IVIM diffusion coefficient (D_s) and ADC increased by 0.60±0.14 ×10⁻³ mm²/s (p<0.0001) and 0.50±0.23 ×10⁻³ mm²/s (p<0.0001), respectively, which correspond to effect sizes of d=4.8 and 2.3 and percent changes of 78±21% and 60±36%, respectively (Table 1 and Figure 1). Both IVIM diffusion coefficient (D_s) and ADC were increased in the ischemic vs. control femoral head in all 11 piglets (Table 2 and Figure 2).

Subtracted CE-MRI images and ADC and IVIM parameter maps for two representative piglets are shown in Figure 3. In these examples, the diffusion parameters ADC and IVIM diffusion coefficient (D_s) were clearly increased (yellow vs. red color) in the SOCs of the ischemic vs. control femoral heads. Conversely, the perfusion fraction (f) and perfusion flux (f*D_f) were considerably decreased in the SOCs of the ischemic vs. control femoral heads. The perfusion coefficient (D_f), on the other hand, was varied in its response, increasing in one case and decreasing in the other.

Figure 3:

Subtracted CE-MRI images, ADC maps, and IVIM parameter maps (diffusion coefficient: D_s; perfusion coefficient: D_f; perfusion fraction: f; and perfusion flux: f*D_f) of the femoral heads of two representative piglets (a and b). CE-MRI confirmed induction of complete femoral head ischemia in the operated hip (yellow arrows). ADC and IVIM diffusion coefficient (D_s) increased in the SOCs (outlined; color overlay) of the ischemic vs. control femoral heads. In contrast, IVIM perfusion fraction (f) and perfusion flux (f*D_f) decreased in the SOCs of the ischemic vs. control femoral heads. IVIM perfusion coefficient (D_f) had variable response, increasing in the ischemic SOC of piglet a and decreasing in the ischemic SOC of piglet b.

IVIM parameters extracted from ROIs independently drawn by the two observers yielded nearly identical results (ICC(2,1)>0.9), confirming the reliability of the ROI definitions.

DISCUSSION:

Our findings obtained from a piglet model of LCPD demonstrate that IVIM is sensitive in detecting bone marrow ischemia of the developing femoral head without the use of an exogenous contrast agent. In particular, IVIM can simultaneously measure a decrease in femoral head perfusion (due to ischemia) and an increase in water diffusion (which is known to result from ischemic injury to the femoral head^10–12); thus, IVIM can assess both the perfusion status of and the severity of necrotic injury to the femoral head. These findings support that IVIM may be a suitable non-contrast-enhanced approach for early detection and monitoring of LCPD to inform treatment decisions.

We found an overall decrease in the IVIM perfusion fraction (f) and flux (f*D_f) in the ischemic vs. control femoral heads, as hypothesized. In particular, the perfusion fraction (f) had the greatest effect size to detect ischemia. This is consistent with decreased fast diffusion signal due to the loss of blood flow in the ischemic bone marrow, which would cause the perfusion fraction (f) to decrease (as occurs in stroke²²). We found that the perfusion fraction (f) decreased ischemic vs. control femoral heads in 9/11 piglets, with differences in f value ranging from 0.032 to 0.093 (18% to 75% decrease). In the two cases where there was negligible decrease in the value of f (0.001 and 0.004), the measured perfusion fraction (f) in the control femoral head was relatively low (0.039 and 0.067, compared to the average control f value of 0.109). Thus, while on average the IVIM method is detecting the hypothesized decrease in perfusion fraction (f) due to ischemia, the method also requires robust measurement of normal femoral head perfusion. We did not find the perfusion coefficient (D_f) to change in a consistent way across the 11 pigs. The perfusion coefficient (D_f) is challenging to measure, particularly if perfusion fraction (f) is very low (as with ischemia), and thus we would not expect this to be a reliable measure of femoral head perfusion. However, given the potential relationship between perfusion fraction (f) and perfusion coefficient (D_f) in the parameter fitting, the perfusion flux (f*D_f) provides a potential alternative measure that can capture changes due to perfusion. We found that perfusion flux (f*D_f) decreased in 10/11 pigs, but it increased in one pig. In this one case, the perfusion fraction (f) was relatively high in both the control and ischemic femoral heads (0.184 and 0.152, respectively), which, combined with a detected increase in perfusion coefficient (D_f) in the ischemic femoral head, led to the measured increase in perfusion flux (f*D_f). Thus, reliability of the measurement of both perfusion fraction (f) and perfusion coefficient (D_f) requires further technical consideration for the method to be used routinely. However, overall, these findings provide key evidence that IVIM is a potentially feasible approach to detect femoral head ischemia and, by extension, other perfusion changes to the bone marrow of the femoral head (such as revascularization).

The sensitivity of IVIM in detecting femoral head ischemia also provides evidence that IVIM can detect normal bone marrow perfusion in the femoral head. On average, we measured the perfusion fraction (f) to be 0.109±0.043 in the SOC of the control femoral heads, which is similar to the f values measured in prior vertebral bone marrow perfusion studies using IVIM. For example: using a similar fat-suppressed RESOLVE-DWI sequence, Lasbleiz et al. measured an average vertebral bone marrow f value of 0.145±0.054 in healthy adults;³² and healthy adult studies by Marchand et al. and Liu et al. measured average vertebral bone marrow f values of 0.140±0.06 and 0.142±0.011, respectively.^25,33 Meng et al. evaluated IVIM perfusion of the femoral head in children 6-24 months old with developmental dysplasia of the hip and measured a high average perfusion fraction (f) of 0.521±0.090 in the control femoral heads.²⁶ This discrepancy with our result may be due to either differences in the subject population and/or the IVIM parameter measurement approach. Segmented fitting approaches, such as the analytical segmented (AS) approach utilized in this work, have been shown to provide improved accuracy compared to the bi-exponential fitting method (as used by Meng et al.), particularly for low-perfused and low-SNR tissues (such as bone marrow).^23,30 As another basis of comparison, Kubo et al. measured femoral head blood volume in healthy adults using positron emission tomography to range from 1.67 to 6.03 mL/100g.³⁴ For our study, if we assume an MR-visible water volume fraction of 35% (consistent with that of hematopoietic bone marrow³⁵) in the SOC, then, using the calculation previously described by Le Bihan et al.,³⁶ our average f value of 0.109 corresponds to a femoral head blood volume of 3.8 mL/100g, which is similar to that measured by Kubo et al. Collectively, our measurements of perfusion fraction (f) are in keeping with prior reports, which further corroborates that IVIM can measure bone marrow perfusion in the femoral head.

In addition to IVIM-detected perfusion changes, we also found that ADC and the IVIM diffusion coefficient (D_s) were sensitive in detecting an increase in diffusion in the ischemic vs. control femoral heads. In particular, we found that the increase in IVIM diffusion coefficient D_s (0.60±0.14 ×10⁻³ mm²/s) was more pronounced than the increase in ADC (0.50±0.23 ×10⁻³ mm²/s). Our ADC results at one-week post-surgery agree with a prior study by Menezes et al.,¹¹ who reported an increase in ADC from 0.96±0.12 ×10⁻³ mm²/s to 1.44±0.13 ×10⁻³ mm²/s one week following induction of ischemia in the femoral head in the LCPD piglet model. Menezes et al. attributed this increase in ADC to the initial microstructural destruction caused by femoral head ischemia. We speculate that the smaller increase in ADC vs. D_s that we observed is due to an artificial increase in the control femoral head ADC value on account of the contribution of the perfusion signal to the low b-value diffusion-weighted images used in the parameter fitting. On the other hand, the IVIM diffusion coefficient (D_s) is measured only at high b-values where the effect of perfusion is negligible. These findings demonstrate a strength of IVIM in providing complementary information on both femoral head perfusion status and diffusion changes in response to ischemic injury within a single acquisition.

While our findings demonstrate the ability of IVIM to detect ischemia and perfusion of the femoral head, technical challenges need to be addressed before clinical introduction of this technique. These challenges, which are common to diffusion-weighted EPI sequences, include low spatial resolution and sensitivity to distortions and artifacts. While readout-segmented EPI acquisitions (such as the RESOLVE sequence used in this study) help reduce these artifacts,³⁷ they also require longer scan times that might not be suitable for imaging pediatric patients. Thus, further technical improvements using innovations such as simultaneous multi-slice imaging and/or denoising techniques^29,38 are likely needed to make application of IVIM in pediatric patients with LCPD feasible. Despite these challenges, this study supports the unique utility of IVIM to simultaneously measure both perfusion and diffusion in the femoral head and motivates further investigation and development of this technique to assess bone marrow perfusion.

Our study has several limitations. First, our sample size is small, but the consistency of our findings across the 11 piglets provides sufficient evidence to support that IVIM is sensitive in detecting changes in both perfusion and diffusion following ischemic injury to the developing femoral head. Second, we used the perfused, contralateral femoral head as a control. While this is a reasonable comparison, ideally the IVIM changes to the SOC would be studied longitudinally, for example, before and after surgical induction of ischemia. Third, we only evaluated the piglets at one time point (one week following onset of ischemia), which corresponds to the early, avascular stage of LCPD. It would also be of interest, in future work, to confirm that IVIM is also sensitive in detecting reperfusion at later time points after initiation of femoral head revascularization and repair. Lastly, IVIM was only evaluated at one central imaging slice. For this validation study, we focused on one slice to reduce potential confounding effects due to partial volume averaging in the small piglet femoral heads.

In conclusion, IVIM is sensitive in detecting bone marrow ischemia of the femoral head in a piglet model of LCPD. IVIM thus has the potential to be used as a non-contrast-enhanced approach to simultaneously assess both bone marrow perfusion and diffusion in patients with LCPD and other ischemic bone and joint disorders. This may provide new opportunities for early detection, staging, monitoring, and serial imaging of LCPD to inform treatment decisions and advance understanding of the disease process.