Compositional evaluation of lesion and parent bone in patients with juvenile osteochondritis dissecans of the knee using T₂* mapping

Abstract

Juvenile osteochondritis dissecans (JOCD) lesions contain cartilaginous, fibrous and osseous tissues which are difficult to distinguish with clinical, morphological magnetic resonance imaging (MRI). Quantitative T₂* mapping has earlier been used to evaluate microstructure and composition of all aforementioned tissues as well as bone mineral density. However, the ability of T₂* mapping to detect changes in tissue composition between different JOCD lesion regions, different disease stages, and between stable and unstable lesions has not been demonstrated. This study analyzed morphological and T₂* MRI data from 25 patients (median age, 12.1 years) with 34 JOCD-affected and 13 healthy knees. Each lesion was assigned a stage reflecting the natural history of JOCD, with stages I and IV representing early and healed lesion, respectively. T₂* values were evaluated within the progeny lesion, interface and parent bone of each lesion and in the control bone region. T₂* was negatively correlated with JOCD stage in progeny lesion (ρ = −0.871; p < 0.001) and interface regions (ρ = −0.649; p < 0.001). Stage IV progeny showed significantly lower T₂* than control bone (p = 0.028). T₂* was significantly lower in parent bone than in control bone of patients with stable lesions (p = 0.009), but not in patients with unstable lesions (p = 0.14). Clinical significance: T₂* mapping enables differentiation between different stages of JOCD and quantitative measurement of the ossification degree in progeny lesion and interface. The observed T₂* decrease in healed and stable lesions may indicate increased bone density as a result of the active repair process. T₂* mapping provides quantitative information about JOCD lesion composition.

1 |. INTRODUCTION

Juvenile osteochondritis dissecans (JOCD) is a developmental joint disorder commonly manifested as an osteochondral lesion of the distal femoral condyle,^1–4 affecting 6–10 of 100,000 children aged 6–19 years.^5,6 Comparative studies suggest a similar pathogenesis in humans and animals with the disease onset related to ischemic necrosis of epiphyseal (subarticular) cartilage and subsequent failure of endochondral ossification.^7–9 Over time, the necrotic epiphyseal cartilage (aka “progeny lesion”) may either heal by becoming a part of subchondral “parent” bone or progress to clefting and eventual creation of a bony fragment.^1–3 Although the lesion ossification has an important role in the pathogenesis,² the natural history of JOCD is still not entirely understood.¹⁰

The majority of previous magnetic resonance imaging (MRI) studies focused on the morphological assessment of JOCD lesions with particular attention to describe status of the underlying articular cartage and the main three JOCD regions: the progeny lesion, the parent bone and the interface between them.^1–3,10,11 Clinical, morphological MRI can differentiate between stable and unstable lesions, and help guide JOCD management.⁴ The interface between progeny lesions and parent bone plays an important role in the evaluation of lesion stability with the presence of multiple cysts or one large cyst suggesting unstable JOCD lesion.⁴ Unstable, symptomatic lesions are treated surgically,¹² while conservative treatment for 6–12 months is recommended for stable lesions.^13,14 Unfortunately, conservative treatment of stable lesions fails in up to 50% of patients.^13,15–17 Although predictors such as age, lesion size, and presence of cystic changes in lesion seem to be associated with treatment success,¹⁶ it is still not possible to reliably predict which lesions will heal and which not. Therefore there is a demand for new, noninvasive MRI biomarkers that are able to predict which lesions will heal and which will become unstable before the morphological signs of instability are visible on clinical, morphological MR images. Noninvasive MRI biomarkers enabling reliable prediction of treatment outcome for conservatively treated stable JOCD lesions (i.e., knee immobilization, limited weight-bearing, activity restriction) may significantly shorten the recommended 6–12 months of nonoperative treatment^13,14 for lesions that are unlikely to benefit from such therapy, and thus improve clinical management of JOCD patients.

Clinical, morphological MRI relies on spin echo sequences with long (≥20 ms) echo times (TE) that cannot unambiguously distinguish between the cartilaginous, fibrous and osseous tissues found within JOCD lesions in previous histological studies.^18–20 The novel concept of short TE, morphological MRI with CT-like contrast was recently introduced to improve the visualization of osseous tissues in JOCD lesions.² Additionally, these MRI techniques were used for staging of JOCD lesions based on depiction of osseous tissue in the progeny lesion and presence of osseous bridging to the interface.² Unfortunately, clinical MRI and short TE, morphological MRI with CT-like contrast are not able to provide information about the structural composition of osseous tissue in JOCD lesions. MRI techniques able to evaluate structure and composition of tissues within JOCD lesions could, therefore, provide new insights into the process of lesion healing.

Here, we propose T₂* mapping for the compositional evaluation of osseous and other tissues within JOCD lesions. T₂* is the transverse relaxation time constant that characterizes how fast the MR signal (i.e., the transverse component of magnetization) decays in the tissue in the presence of local static magnetic field inhomogeneities. T₂* values reflect the interactions between the magnetic moments of water and the surrounding macromolecules in the given tissue and therefore provides indirect measurement of tissue microstructure and composition.^21,22 Quantitative T₂* mapping has been successfully used to evaluate microstructure and composition of the articular cartilage,^23,24 fibrocartilage,²⁵ and trabecular bone.^26–29 Previously reported T₂* values in cartilage (~23 ms) ^23,24 and fibrocartilage (~19 ms) ²⁵ are much higher compared to T₂* in trabecular bone (~4 ms),^30–32 T₂* mapping should therefore allow differentiation between cartilaginous and osseous regions within JOCD lesions. Furthermore, magnetic susceptibility differences between bone trabeculae and marrow impose static field inhomogeneities and causes shortening of T₂* values. T₂* values in the bone marrow thus provide indirect information about the structure and density of surrounding trabecular architecture.^33–35 This unique feature of T₂* mapping was used in bone marrow studies that showed significantly longer T₂* in patients with osteoporosis compared to control subjects and demonstrated negative correlation between T₂* values and bone mineral density in femur, calcaneus and lumbar spine.^26–29 Noninvasive T₂* mapping may therefore provide a useful continuous metric for differentiation and quantitative evaluation of cartilaginous and osseous tissue components as well as indirect assessment of trabecular bone quality in JOCD lesions.

Considering the difference in T₂* values between articular cartilage ^23,24 and trabecular bone ^26–29 reported in previous studies, the goal of the present study was to determine if T₂* mapping allows noninvasive, quantitative assessment of all lesion tissues (i.e., cartilaginous and osseous) and evaluation of their structure and composition in JOCD patients. Therefore, we retrospectively evaluated participants with JOCD lesions in distal femoral condyle to assess ability of quantitative T₂* mapping to detect differences in tissue composition between: (i) different regions within JOCD lesions, (ii) different JOCD stages, and (iii) between stable and unstable JOCD lesions.

2 |. METHODS

2.1 |. Study population

This retrospective observational cohort study (level of evidence: 3) adhered to Health Insurance Portability and Accountability Act and was approved by the institutional review board that waived the need for informed consent. Between December 2015 and March 2020, a total of 37 young patients with initial diagnosis of JOCD underwent 3T MRI including T₂* mapping at our institution. From the cohort of 37 patients, two patients were excluded because their lesions were diagnosed as ossification variants.³⁶ Four patients were excluded because of the absence of an open growth plate in the distal femur. Six patients were excluded due to the prior surgery in the evaluated knee. Finally, 25 patients with JOCD of the knee were included into our evaluations. Exclusion criteria and the corresponding numbers of excluded participants are summarized in Figure 1. None of patients showed signs of knee arthritis, malignancy, or severe MRI artifacts. Due to the relatively high prevalence of bilateral JOCD lesions, both knees were imaged. Inclusion criteria for healthy, control knees were absence of clinical symptoms in the knee (e.g. pain, clicking, locking), no JOCD lesion, and no abnormal MRI findings (e.g. bone marrow lesions, cartilage, or meniscus abnormalities).

FIGURE 1.

Flowchart of study population

2.2 |. MRI protocol

All images were acquired on a whole-body 3T MRI system (Prisma Fit; Siemens Healthcare) using a single-channel transmit, 15-channel receive phased-array knee coil (Quality Electrodynamics). The protocol included morphological T₂-weighted, T₁-weighted, and proton density-weighted turbo-spin echo sequences with and without fat suppression. For the evaluation of T₂* relaxation times, a multislice multiecho gradient recalled echo (GRE) sequence was acquired with following parameters: repetition time = 1150 ms; six TEs = 2.6, 5.6, 8.5, 11.5, 14.5 and 17.4 ms; in plane resolution = 0.43 × 0.43 mm²; slice thickness = 2.0 mm; acquisition time of about 6 min (depending on the number of acquired slices). The TEs were selected as a compromise between the signal-to-noise ratio in the trabecular bone marrow, amount of data points for fitting of short T₂* in bone marrow, longest TE close to expected T₂* of cartilage and the total acquisition time below 7 min. All six TEs were acquired with a single multiecho GRE sequence. The short echo images of the T₂* mapping sequence are routinely clinically reviewed by the radiologists at our institution for the evaluation of osseous components within lesion and assignment of JOCD stage² described in the following section. Acquisition parameters of all MRI sequences are detailed in Table 1.

TABLE 1.

Acquisition parameters of MRI sequences

Acquisition	Axial FS	Sagittal	Sagittal FS	Coronal FS	Coronal	Coronal 2D
Parameter	T2w TSE	PDw TSE	PDw TSE	T2w TSE	T1w TSE	T2* mapping
Repetition time (ms)	3860	1810	2000	3860	786	1150
Echo time (ms)	36	37	32	47	9.3	2.6, 5.6, 8.5, 11.5, 14.5, 17.4
Flip angle (degrees)	150	150	150	150	150	60
Pixel BW (Hz/pix)	195	240	240	200	200	405
Field of view (mm2)	130 × 130	110 × 110	110 × 110	110 × 110	110 × 110	150 × 150
Matrix size	256 × 233	256 × 218	256 × 218	320 × 224	320 × 224	352 × 352
Resolution# (mm2)	0.51 × 0.51	0.43 × 0.43	0.43 × 0.43	0.34 × 0.34	0.34 × 0.34	0.43 × 0.43
Slice thickness (mm)	3.0	3.0	3.0	3.0	3.0	2.0
Number of slices	31	31	31	29	29	30–39
Echo train length	11	5	5	11	2	6
Number of averages	1	1	2	2	1	1
Freq.-encoding direction	A-P	A-P	A-P	F-H	F-H	R-L
Imaging time (min)	1:17	1:21	2:54	2:22	1:01	5:24–6:57

Abbreviations: A-P, anterior-posterior; BW, bandwidth; F-H, feet-head; Freq.-encod., frequency-encoding; FS, fat suppression; PDw, proton density-weighted; R-L, right-left; T₁w, T1-weighted; T₂w, T2-weighted; TSE, turbo spin echo.

^#Reconstructed pixel size is reported.

2.3 |. Lesion stability and staging

Two musculoskeletal radiologists (S.M. and J.M.E., both with 15 years of experience) independently evaluated the lesion stability on clinical, morphological MR images by using the morphologic features of instability as previously described by Kijowski et al.⁴ An unstable JOCD lesion was defined as the presence of articular cartilage fracture and/or the presence of four secondary signs: a rim of fluid signal intensity, a second outer rim of low T₂ signal intensity, multiple breaks in the subchondral plate, and a large or multiple cyst(s).⁴ After independent evaluations of lesion stability, all lesions that were evaluated differently were reviewed and the two radiologists came to a consensus in each JOCD lesion.

Two fellowship-trained musculoskeletal radiologists (S.M. and J.M.E.) without prior access to the patients’ clinical history and T₂* maps independently assigned each lesion to one JOCD stage according to previously described staging system ² based on qualitative depiction of osseous tissues in the progeny lesion and its interface to parent bone on morphological, short TE GRE images with CT-like contrast. This system divides JOCD lesions into 5 stages: (I) epiphyseal cartilage lesion with delay in endochondral ossification (Figure 2A); (II) peripheral ossification of the progeny (Figure 2B); (III) partial ossification of the progeny with partial osseous bridging (Figure 2C); (IV) completely ossified progeny with complete osseous bridging—healed or almost healed lesion (Figure 2D); or (V) not-healed detached lesion. Lesions that showed evidence of two different stages were assigned a stage representing the majority of lesion volume. After independent staging, the two radiologists performed a review of all lesions they staged differently and came to a consensus in each case.

FIGURE 2.

A qualitative depiction of osseous tissues in JOCD lesions (between arrows) on the morphological, short echo time GRE images with CT-like contrast showing different stages of disease. (a) A 12-year-old boy with a stage I, cartilaginous-only lesion. (b) A 12-year-old girl with a stage II lesion showing ossification of the progeny rim (arrow heads). (c) A 14-year-old boy with a stage III, predominantly osseous progeny with a partial osseous bridging to parent bone (arrow head). (d) A 13-year-old girl with a stage IV, osseous, healed lesion. All lesions are located on the medial femoral condyle [Color figure can be viewed at wileyonlinelibrary.com]

2.4 |. Segmentations and quantitative evaluations

Manual 3D segmentations were performed on the T₂*-weighted images with the shortest TE using ITK-SNAP,³⁷ with readers having access to morphological MRI (Figure 3A,B) but not to the T₂* maps. In JOCD knees, four regions, including three regions within JOCD lesion complex and a control bone region, were segmented as depicted in Figure 3C. The progeny lesion, the parent bone and the interface between them are the three distinct JOCD regions, each having specific morphologic features and a unique role during the disease progression.^1–3,10,11 The progeny lesion region was segmented to encompass the entire progeny area. The interface region was selected between progeny lesion and adjacent trabecular bone. The parent bone region was selected to encompass the sclerotic bone adjacent to the interface region. The control bone region was segmented within trabecular bone of the contralateral, lesion-free femoral condyle. In healthy, JOCD-free knees, only the control bone region was segmented matching the control bone region in JOCD knee of the same patient. To provide one representative T₂* value for healthy control bone that could be compared to T₂* values from all three regions within JOCD lesion (i.e. progeny lesion, interface, parent bone) we attempted to match the size of the control bone region to the combined size of progeny lesion, interface and parent bone regions. All regions were segmented by a radiology resident (C.S.) in consensus with an expert musculoskeletal radiologist (J.M.E.). For the evaluation of interobserver reproducibility, a musculoskeletal MRI physicist (Š.Z., 11 years of experience) independently segmented the regions in 10 randomly selected JOCD knees.

FIGURE 3.

An example of manual segmentations of the four evaluated regions on MR images of a 14-year-old boy with a stable, stage III JOCD lesion on the medial femoral condyle. (a) The T₂-weighted turbo spin echo image with fat suppression depicts the position of progeny lesion (between arrows) and interface as well as hyperintense edema in the parent bone (asterisks). (b) The T₁-weighted turbo spin echo image is showing a parent bone region with the replacement of normal fatty marrow (between arrows). (c) All four evaluated regions were selected on the first echo of the T₂*-weighted MR images. Three regions were part of the JOCD lesion complex: progeny lesion (red), interface (yellow) and parent bone (green). Additionally a control bone region (blue) served as a reference. (d) A higher magnification view of the JOCD lesion area showing progeny lesion, interface and parent bone detail. (e) Segmented regions (white contours) overlaid on the color-coded T₂* map; the color bar represents T₂* values in milliseconds. (f) The color-coded coefficient-of-determination (R²) map with four segmented regions (white contours); the color bar represents R² values [Color figure can be viewed at wileyonlinelibrary.com]

All T₂* maps were calculated on a pixel-wise basis by fitting the MR signal to the model of signal decay using a two-parametric nonlinear least-square fitting with the Levenberg-Marquardt algorithm in Matlab (R2017b; MathWorks) (Figure 3E). Following exponential model was fitted to multiecho data with all TEs contributing equally

$$S=M_0\exp \left(-TE/T_2^{*}\right),$$

where TE is the echo time, S is measured MR signal at a given TE, and the parameters M₀ and T₂* are estimated magnetization in equilibrium and T₂* relaxation time, respectively (Figure 4D–F). The corresponding coefficient of determination (R²) maps were calculated to evaluate the percentage of variation explained by the fitting model (Figure 3F). Median T₂* and R² values were calculated from the progeny lesion, interface, parent bone and control bone 3D regions. Additionally, region volumes were calculated by multiplying a region’s pixel count with the pixel volume of the T₂*-weighted images.

FIGURE 4.

A 12-year-old girl with a stage II unstable JOCD lesion on the medial femoral condyle. (a) The T₂-weighted turbo spin echo images with fat suppression depict the lesion location, a hyperintense area of edema in the parent bone, a fluid-like high signal rim in the interface and a break in the articular cartilage and the subchondral bone plate (arrowhead). (b) Short echo time gradient echo image with CT-like contrast showing a high signal of the progeny rim ossification and a low signal of cartilaginous areas in the progeny lesion and interface. (c) The corresponding color-coded T₂* map with four selected regions (white contours). Please note the parent bone with lower T₂* (blue areas) compared to the control bone region on the opposite condyle which may indicate a decrease in bone density in the parent bone. Additionally, T₂* map reflects the heterogeneous composition of progeny lesion region with high T₂* areas (red) being composed predominantly of cartilaginous tissue and low T₂* areas (green and blue) of osseous tissues. A higher magnification image of the JOCD lesion area is shown in the lower right corner of each image. The color bar represents T₂* values in milliseconds. (d) A plot showing a representative T₂* fit (blue line) of signal intensities (black points) as a function of echo time from a single pixel situated in the cartilaginous tissue in the progeny lesion shown in (a–c) (fitted T₂* = 27.1 ms; goodness of fit (R²) = 0.963). (e) A representative T₂* fit of signal intensities as a function of echo time from a single pixel situated in the osseous area of progeny lesion shown in (a–c) (fitted T₂* = 3.7 ms; R² = 0.959). Please note the faster decay of signal intensities with echo time and therefore shorter T₂* when compared to the fit illustrated in (d) [Color figure can be viewed at wileyonlinelibrary.com]

2.5 |. Statistical methods

Agreement between the two radiologists evaluating stability and stage of JOCD lesions was calculated with the Cohen’s κ. The mixed effects regression models, with age, sex and lesion volume as covariates and adjustment for within-subject variability, were followed by Tukey’s posthoc tests to compare T₂* values between the segmented regions, the JOCD stages (I–IV), the healthy and JOCD knees, and the stable and unstable lesions. The same tests were used to compare the lesion volume between the JOCD stages. p values were adjusted for multiple pair-wise comparisons. Age and lesion volume differences between the stable and unstable lesions were evaluated with independent samples t tests. The inter-reader reproducibility of T₂* evaluations was assessed by the Pearson’s correlation coefficients and the coefficients of variance. Spearman rank correlation coefficients (ρ) were calculated to evaluate the correlations of T₂* and region volume with JOCD stage and patient’s age. Statistical significance was indicated by a p < 0.05. All evaluated data were examined for normal distribution using Box plots and Q-Q plots. Region volumes were normalized by calculating the cube root of volume to ensure the normal distribution of data before further statistical evaluations. Statistical analyses were calculated in R (3.5.1; Foundation for Statistical Computing).

3 |. RESULTS

3.1 |. Patient cohort

A total of 25 patients (16 boys [median age, 13.2 years; interquartile range [IQR], 11.7–14.9 years], 9 girls [median age, 11.6 years; IQR, 9.8–12.7 years]) met all inclusion criteria and were enrolled in this study (Table 2). This cohort presented with 34 JOCD-affected knees and 13 lesion-free knees. From 22 patients with bilateral MRI, nine (41%) had JOCD lesions in both knees. All 34 lesions were divided into four subgroups according to the JOCD staging system.² The staging of lesions by the two independent radiologists resulted in high inter-rater agreement with a Cohen’s κ of 0.838 (95% confidence interval [CI] = 0.689, 0.988) and agreement in 30 out of 34 (88%) lesions.

TABLE 2.

Demographic details of the study group and subgroups

		Sex	Age (years)	Laterality	Location	Symptoms	Lesion stability
Group	Number of knees (%)	Male/female	Median (IQR)	Right/left	MFC/LFC	Yes/no	Stable/Unstab.
JOCD I	5 (10.6%)	3/2	8.8 (7.6–12.0)	3/2	2/3	4/1	5/0
JOCD II	13 (27.7%)	7/6	12.8 (12.0–13.5)	9/4	12/1	13/0	6/7
JOCD III	9 (19.1%)	8/1	12.1 (11.5–14.9)	1/8	8/1	9/0	5/4
JOCD IV	7 (14.9%)	3/4	11.6 (11.4–13.4)	2/5	4/3	3/4	7/0
Control	13 (27.7%)	9/4	11.7 (10.3–14.9)	8/5	N/A	0/13	N/A
Total	47 (100%)	30/17	12.1 (11.5–14.5)	23/24	26/8	29/18	23/11
Stable	11 (23.4%)	7/4	12.1 (11.7–13.1)	5/6	10/1	10/1	11/0
Unstable	11 (23.4%)	8/3	13.2 (12.3–16.5)	5/6	10/1	11/0	0/11

Abbreviations: IQR, interquartile range; JOCD, Juvenile Osteochondritis Dissecans, LFC, lateral femoral condyle; MFC, medial femoral condyle, Unstab., unstable.

Eleven of the 34 JOCD lesions were evaluated as unstable using the morphological MRI criteria of instability.⁴ The evaluation of lesion stability by the two independent radiologists resulted in substantial inter-reader agreement with a Cohen’s κ of 0.656 (95% CI = 0.381, 0.931) and agreement in 29 out of 34 (85%) JOCD lesions. All unstable lesions were surgically treated after the MRI (median interval, 47 days; IQR, 28–100 days). Example morphological MR images and a T₂* map of unstable JOCD lesion are illustrated in the Figure 4. From 23 stable lesions, 11 lesions were selected to match for JOCD stage and other patient characteristics of unstable lesion group (Table 2). No statistically significant difference in age was observed between stable and unstable groups (p = 0.07). All patients from stable lesions group became asymptomatic during conservative treatment (shortest follow-up, 12 months) and three JOCD lesions were confirmed by imaging to heal.

3.2 |. Lesion regions

The comparisons between evaluated regions found significantly higher T₂* values in the progeny lesion than in the parent and control bone regions at JOCD stages I–III (all p ≤ 0.025) (Table 3). At stage IV, progeny lesion T₂* was significantly lower than control bone T₂* (p = 0.028), but not than parent bone T₂* (p = 0.45). This difference is demonstrated on T₂* map in the Figure 5. Additionally, interface T₂* was significantly higher than parent bone T₂* at stages II and III, and control bone T₂* at stage II (all p ≤ 0.006). Although differences were not statistically significant, T₂* was lower in parent bone than in control bone at all JOCD stages (all p ≥ 0.17). Furthermore, control bone T₂* was not significantly different between the control knees (median, 5.0 ms; IQR, 4.6–5.4 ms) and the JOCD knees (median, 4.8 ms; IQR, 4.4–5.3 ms) of the same patients (p = 0.06). Median R² values in all regions were higher than 0.91 indicating reliable T₂* fitting.

TABLE 3.

Comparison between different regions

		T2* (ms)			R 2	Volume (mm3)
Knee type	Region	Median (IQR)	p values		Median (IQR)	Median (IQR)
JOCD Stage I n = 5	Progeny lesion	26.0 (21.6–28.6)	vs. IN, <0.001	vs. PB, <0.001	0.948 (0.943–0.956)	191 (124–230)
	Interface	10.7 (10.6–11.8)	vs. PB, =0.28	vs. CB, =0.62	0.958 (0.956–0.962)	120 (101–146)
	Parent bone	3.9 (3.6–3.9)	vs. CB, =0.98		0.935 (0.920–0.944)	377 (336–610)
	Control bone	4.2 (4.1–5.2)	vs. PL, =0.008		0.935 (0.934–0.939)	723 (537–828)
JOCD Stage II n = 13	Progeny lesion	17.4 (13.8–23.5)	vs. IN, =0.086	vs. PB, <0.001	0.953 (0.948–0.956)	539 (243–801)
	Interface	10.2 (8.7–11.8)	vs. PB, <0.001	vs. CB, <0.001	0.957 (0.943–0.962)	295 (120–346)
	Parent bone	4.3 (3.8–4.5)	vs. CB, =0.93		0.943 (0.935–0.950)	938 (525–1437)
	Control bone	4.8 (4.5–5.2)	vs. PL, <0.001		0.942 (0.937–0.944)	2177 (942–2306)
JOCD Stage III n = 9	Progeny lesion	8.7 (6.7–14.6)	vs. IN, =0.73	vs. PB, =0.001	0.951 (0.941–0.959)	470 (376–739)
	Interface	9.6 (7.6–12.7)	vs. PB, =0.006	vs. CB, =0.06	0.953 (0.945–0.964)	247 (155–323)
	Parent bone	4.2 (3.8–4.5)	vs. CB, =0.99		0.939 (0.919–0.947)	791 (769–1015)
	Control bone	5.1 (4.5–5.5)	vs. PL, =0.025		0.934 (0.925–0.947)	1454 (1127–1910)
JOCD Stage IV n = 7	Progeny lesion	3.5 (3.4–3.9)	vs. IN, =0.84	vs. PB, =0.45	0.936 (0.927–0.939)	216 (174–494)
	Interface	3.6 (3.3–3.8)	vs. PB, =0.34	vs. CB, =0.11	0.927 (0.917–0.937)	64 (54–120)
	Parent bone	4.0 (3.9–4.3)	vs. CB, =0.17		0.933 (0.925–0.942)	424 (240–595)
	Control bone	4.9 (4.4–5.3)	vs. PL, =0.028		0.927 (0.925–0.945)	609 (458–1081)
Control n = 13	Control bone	5.0 (4.6–5.4)	vs. CB in JOCD, =0.06		0.938 (0.922–0.940)	1545 (870–2024)

Abbreviations: CB, control bone; IN, interface; IQR, interquartile range; JOCD, Juvenile Osteochondritis Dissecans; PB, parent bone; PL, progeny lesion; R², coefficient of determination.

FIGURE 5.

A 13-year-old girl with a healed, stage IV JOCD lesion on the medial femoral condyle. (a) The first echo of the T₂*-weighted MR images with CT-like contrast showing osseous, healed lesion (between arrows). (b) The corresponding color-coded T₂* map with four selected regions in white contours: progeny lesion, interface, parent bone, and control bone on opposite condyle. Note the lower T₂* values in the progeny lesion than in the control bone region; the color bar represents T₂* values in milliseconds. (c) The color-coded coefficient-of-determination (R²) map shows high agreement (close to 1) between the measured data and the exponential fit in all four evaluated regions (white contours); the color bar represents dimensionless R² values. (d) A zoomed-in depiction of the JOCD lesion area on T₂* map showing detail of progeny lesion, interface and parent bone; the color bar represents T₂* values in milliseconds [Color figure can be viewed at wileyonlinelibrary.com]

3.3 |. JOCD stages

The comparisons between different JOCD stages showed significantly higher T₂* in progeny lesion at stages I and II than at stages III and IV (all p ≤ 0.011) (Table 4). The interface showed significantly lower T₂* at stage IV compared to stages I–III (all p ≤ 0.046). No significant differences between different JOCD stages were found in parent bone T₂*, control bone T₂*, or volume of progeny lesion (all p ≥ 0.12).

TABLE 4.

Comparisons between different JOCD stages

Lesion stage	Progeny T2*	Interface T2*	Parent bone T2*	Control bone T2*	Progeny volume
I vs. II	0.053	0.95	0.66	0.47	0.17
I vs. III	<0.001	0.75	0.70	0.12	0.40
I vs. IV	<0.001	0.003	0.49	0.43	0.66
II vs. III	0.011	0.83	0.99	0.67	0.94
II vs. IV	<0.001	0.020	0.99	0.99	0.72
III vs. IV	0.35	0.046	0.98	0.73	0.97

Note: Data are p values. Bold numbers indicate statistically significant differences.

The Spearman rank correlations showed a significant negative association between the JOCD stage and T₂* in the progeny lesion (ρ = −0.871; 95% CI = −0.936, −0.732; p < 0.001) and in the interface (ρ = −0.649; 95% CI = −0.834, −0.364; p < 0.001) (Figure 6). While no significant correlation was observed between the age and T₂* in all regions, the age was significantly correlated with the volume of progeny lesion (ρ = 0.534; 95% CI = 0.123, 0.753; p = 0.001), interface (ρ = 0.477; 95% CI = 0.165, 0.704; p = 0.004), and parent bone (ρ = 0.415; 95% CI = 0.106, 0.681; p = 0.014). No correlation was found between the age and the JOCD stage (ρ = 0.081; 95% CI = −0.331, 0.477; p = 0.65) (Figure 6). All evaluated Spearman rank correlations are listed in the Table 5.

FIGURE 6.

Regression analysis plots. (a) A significant negative spearman rank correlation (ρ) was observed between the progeny lesion T₂* and the JOCD stage (ρ = −0.871; 95% confidence interval = −0.936, −0.732; p < 0.001). (b) A significant negative correlation was found between the interface T₂* and the JOCD stage (ρ = −0.649; 95% confidence interval = −0.834, −0.346; p < 0.001). (c) A significant positive correlation was observed between the progeny volume and the patient’s age (ρ = 0.534; 95% confidence interval = 0.123, 0.753; p = 0.001). The high median T₂* values (>15 ms) in the interface at JOCD stages II and III can be explained by the presence of fluid in the interface which was detected with clinical, morphological MRI in eight unstable lesions. (d) The spearman rank correlation did not show any significant association between the patient’s age and the JOCD stage (ρ = 0.081; 95% confidence interval = −0.331, 0.477; p = 0.65) [Color figure can be viewed at wileyonlinelibrary.com]

TABLE 5.

Spearman rank correlation analyses

	Region
	Progeny Lesion	Interface	Parent Bone	Control Bone
T2* values vs. JOCD stage
Spearman	−0.871	−0.649	0.107	0.210
p value	<0.001	<0.001	0.55	0.23
95% CI	−0.936/−0.732	−0.834/−0.346	−0.252/0.427	−0.161/0.576
T2* values vs. age
Spearman	−0.111	0.010	−0.040	−0.177
p value	0.53	0.58	0.83	0.32
95% CI	−0.441/0.260	−0.267/0.436	−0.369/0.308	−0.525/0.197
Region volume vs. JOCD stage
Spearman	0.109	−0.231	−0.105
p value	0.54	0.19	0.55	N/A
95% CI	−0.297/0.475	−0.566/0.134	−0.475/0.279
Region volume vs. age
Spearman	0.534	0.477	0.415
p value	0.001	0.004	0.014	N/A
95% CI	0.123/0.753	0.165/0.704	0.106/0.681

Note: Bold numbers indicate statistically significant correlations.

Abbreviations: CI, confidence interval; JOCD, Juvenile Osteochondritis Dissecans.

3.4 |. Stable versus unstable JOCD lesions

Although T₂* values in the interface and parent bone were higher in unstable than in stable JOCD lesions, no significant differences were observed in any of evaluated regions between unstable and stable lesions (all p ≥ 0.27) (Figure 7). We found significantly lower T₂* in parent bone than in control bone of patients with stable lesions (p = 0.009). However, no statistically significant difference was observed between the parent bone and control bone of patients with unstable lesions (p = 0.14) (Figure 7). Our results showed that the volume of progeny lesion was significantly higher in unstable than in stable JOCD lesions (p = 0.002) (Table 6).

FIGURE 7.

Box plots of median T₂* from progeny lesion, interface, parent bone and control bone regions of eleven stable and eleven unstable JOCD lesions. Data points for each group and region are shown next to the corresponding box plot. The mixed effects regression models, with age, sex and progeny volume as covariates and adjustment for within-subject variability found significantly lower T₂* values in parent bone than in control bone in stable lesions (p = 0.009). The high variability of T2* values in the interface of unstable lesion, when compared to stable lesions, is likely due to the presence of fluid detected in 8 of 9 unstable lesions. In each box plot, the cross represents the mean T₂* value and the central horizontal line the median T₂* value of the evaluated region. The upper and lower whiskers extend to the maximum and minimum T₂* values in the region, respectively. The upper and lower borders of the box represent the third quartile (i.e., 75th percentile) and the first quartile (i.e., 25th percentile) of the T₂* data within the region, respectively [Color figure can be viewed at wileyonlinelibrary.com]

TABLE 6.

Comparison between stable and unstable JOCD lesions

	T2* (ms)				Volume (mm3)
	Median (IQR)				Median (IQR)
Lesion type	Progeny lesion	Interface	Parent bone	Control bone	Progeny lesion
Stable	10.8 (6.7–17.2)	5.6 (5.5–8.9)	3.7 (3.3–3.8)	5.2 (4.6–5.3)	199 (184–376)
Unstable	14.7 (13.3–19.2)	12.7 (10.2–21.7)	4.5 (4.1–4.6)	4.8 (4.5–5.2)	801 (458–1022)
p value	0.29	0.27	0.74	0.98	0.002

Note: Bold number indicate statistically significant correlations.

Abbreviation: IQR, interquartile range.

3.5 |. Inter-reader repeatability

High inter-reader repeatability was found for T₂* evaluations. The Pearson’s correlation coefficients between the two readers were 0.997 in progeny lesion (95% CI = 0.987, 0.999; p < 0.001), 0.991 in interface (95% CI = 0.963, 0.998; p < 0.001), 0.885 in parent bone (95% CI = 0.576, 0.973; p < 0.001), and 0.983 in control bone (95% CI = 0.927, 0.996; p < 0.001). The mean coefficients of variation between the two readers were 5.7% in progeny lesion, 6.9% in interface, 4.7% in parent bone, and 2.0% in control bone.

4 |. DISCUSSION

This is the first study to investigate the ability of T₂* mapping to detect differences in tissue composition between different lesion regions, different disease stages, as well as between stable and unstable lesions of JOCD patients. We demonstrated that T₂* mapping allows quantitative evaluation of osseous, fibrous, and cartilaginous tissues in JOCD lesions and can therefore measure the degree of ossification in progeny and interface, which are important factors in lesion healing but are difficult to evaluate with morphological MRI. Our findings suggest that T₂* mapping can detect increased bone density in progeny lesion and parent bone which may indicate an active process of lesion repair. This study demonstrates that the addition of T₂* mapping to the standard clinical MRI protocol provides the noninvasive imaging tool for quantitative tissue characterization, otherwise only attainable by histology. T₂* mapping is a promising method with a potential to improve clinical management of JOCD patients.

We found a progressively decreasing T₂* values with increasing JOCD stage in progeny lesion and interface which is due to the gradual ossification of lesion. The T₂* of progeny lesion at stage I was high and similar to previously reported T₂* of articular cartilage^23,24 and fibrocartilage,²⁵ while the progenyT₂* at stage IV was low and similar to the T₂* of trabecular bone.^30–32 These results emphasize the role of progressive ossification in lesion healing in the natural history of JOCD,^2,7–9,18 and confirm the previously proposed JOCD staging system based on qualitative evaluation of osseous tissues on short TE GRE images with CT-like contrast.² TheT₂* mapping allows quantitative evaluation of all tissues present in JOCD lesions^18–20 while allowing the differentiation between osseous and fibrous tissues, which are difficult to distinguish on standard clinical, morphological MRI sequences with long (≥20 ms) TE. Quantitative measurement of the degree of tissue ossification using T₂* mapping may provide important insights into JOCD lesion healing.

Our results showed significantly lower T₂* in stage IV progeny lesion, signaling a different composition of osseous progeny compared to the healthy, control bone. Significantly lower T₂* values were also found in the parent bone of stable (stage II and III) lesions, but not in the parent bone of unstable (stages II and III) lesions when compared to control bone. All patients with unstable lesions received surgery while all patients with stable lesions became asymptomatic (shortest follow-up of 12 months) and three lesions were confirmed by imaging to heal. Previous in vivo studies reported significant negative correlations of T₂* with bone density evaluated by dual energy X-ray absorptiometry in different skeletal locations including calcaneus, lumbar spine and femoral neck.^26–29 Lower T₂* values probably indicate increased bone density in progeny lesion and parent bone of patients with stable, healing JOCD lesions. This explanation is supported by histological studies that demonstrated increased bone formation, bone resorption and osteoid accumulation in stable JOCD lesions, which are typical characteristics of active tissue repair.^19,38–40 Correspondingly, the absence of significantly lower T₂* in parent bone of unstable lesions may suggest an absence of active healing processes. Our findings suggest that T₂* mapping can detect increased bone density in progeny lesion and parent bone which may indicate active lesion repair, however, future studies are necessary to confirm this relationship. If confirmed, T₂* mapping may provide a new, noninvasive biomarker of lesion repair and thus inform treatment decision in JOCD patients.

We observed almost perfect inter-rater agreement in lesion staging and substantial agreement in the evaluation of lesion instability between two independent radiologists. While the inter-reader agreement for staging of JOCD lesions was not previously evaluated, relatively low interrater reliability of lesion instability assessment was reported.^11,41 Additionally, high interobserver repeatability of T₂* evaluations was found in all evaluated regions. Furthermore, all regions showed median R² values higher than 0.91 indicating that at least 91% percent of the variation in T₂* data can be explained by the fitting model, thus suggesting reliable T₂* mapping in all evaluated tissues. These findings demonstrate that T₂* mapping is reproducible, reader independent method that can serve as an objective outcome measure for the quantitative assessment of JOCD lesions.

While factors such as age, symptoms and lesion location were very similar between the patient groups with stable and unstable JOCD lesions, we observed significantly larger volume of progeny lesion in patients with unstable than with stable lesions. In previous study, lesion width and patient age have been reported to be among the most significant factors for the prediction of lesion healing.¹⁶ Additionally, in the present study, T₂* in the interface tended to be higher in unstable than in stable lesions which is likely due to the presence of fluid in the interface which was detected with clinical, morphological MRI in eight unstable lesions. All unstable lesions were found at stages II and III but not at stage I, which may suggest that the progeny lesion has to be at least partially ossified before the instability can develop. Although a positive association between the JOCD stage and age was reported previously,² we did not observe any significant correlation between these parameters in this study. This could be due to the absence of stage V lesions and the exclusion of patients with a closed femoral growth plate in the current study, or because of the small number of stage I (n = 1) and stage IV (n = 2) lesions in the previous report.² However, we found positive correlation between the age and the volume of progeny lesion, interface and parent bone, indicating that older patients tend to have larger JOCD lesions.

This study has several limitations. First, 34 JOCD lesions is a relatively small sample size. This is mainly due to single-center study design and the rarity of this disease. Second, groups of patients with different JOCD stage were not matched for age, sex or number of subjects due to the retrospective nature of the study. However, no significant correlation was observed between the T₂* and patient’s age in any region, including control bone. Third, lesion-free asymptomatic contralateral knees of JOCD patients without abnormalities on morphological MRI were used as controls. While the biomechanics of control knees might have been altered due to the symptoms in JOCD-affected knees, this approach enabled comparisons between the perfectly age- and sex-matched groups. Fourth, this study doesn’t compare T₂* results to the gold-standard histopathological analysis or to the dual energy X-ray absorptiometry. This is due to retrospective nature of this study and the fact that these methods are not routinely used for the evaluation of pediatric JOCD patients. Finally, since TEs used in this study are neither in phase nor out of phase, a combination of water and fat MR signals was used in the present T₂* analyses of bone marrow. Possible differences in fat and water fractions between the patients might therefore have contributed to the T₂* differences observed in this study. Future studies using multiecho GRE sequence optimized for water-fat separation^22,31,42 are needed to evaluate possible changes in proton density fat fraction during the process of JOCD lesion healing.

In conclusion, T₂* mapping of JOCD lesions allows excellent inter-reader reproducibility, enables evaluation of osseous, fibrous, and cartilaginous tissues in JOCD lesions, and thus provides quantitative measurement of degree of ossification in progeny lesion and interface regions. Furthermore, T₂* results suggest different quality of osseous tissue in recently healed progeny lesions and in parent bone of stable lesions when compared to healthy trabecular bone in JOCD patients. T₂* is a potential imaging biomarker of lesion healing that may be useful in planning and monitoring nonsurgical treatment of JOCD patients. Future studies comparing T₂* results with histology or dual energy X-ray absorptiometry as well as longitudinal T₂* studies are warranted to further validate these promising results and to assess the potential of T₂* in lesion and parent bone as a predictor of JOCD healing.