Introduction

Legg-Calve´-Perthes disease (LCP) is an idiopathic
avascular necrosis of the growing femoral head that
remains a challenge with regard to aetiology, early
prognosis and treatment. According to some authors,
symptoms occur after multiple infarcts because of in-
sufficient revascularisation of an already vascular-
compromised femoral epiphysis [1, 2]. Others, based on
bone scintigraphy, suggest that a vascular insult can
lead to complete interruption of the epiphyseal blood

supply [3]. In any case, subsequent necrosis and bone
resorption are followed by reparative processes and
revascularisation.

The ability of bone scintigraphy to demonstrate

bone perfusion provides a sensitive and specific tech-
nique for the diagnosis of LCP disease. Furthermore,
bone scintigraphy can serve as a prognostic tool by
showing the early revascularisation changes before
changes observed on plain radiographs [3]. Different
pathways of revascularisation have been described with
serial bone scintigraphy [3, 4]. Conway and Dias have

O R I G I N A L A R T I C L E

Pediatr Radiol (2002) 32: 580–585
DOI 10.1007/s00247-002-0732-5

Sylvie Lamer
Sophie Dorgeret
Abdeslam Khairouni
Keyvan Mazda
Pierre-Yves Brillet
Eric Bacheville
Juliette Bloch
Georges F. Pennec¸ot
Max Hassan
Guy H. Sebag

Femoral head vascularisation
in Legg-Calve´-Perthes disease: comparison
of dynamic gadolinium-enhanced subtraction
MRI with bone scintigraphy

Received: 12 October 2001
Accepted: 26 March 2002
Published online: 14 June 2002

Springer-Verlag 2002

S. Lamer Æ S. Dorgeret Æ P.-Y. Brillet
M. Hassan Æ G.H. Sebag (

&)

Department of Paediatric Radiology,
Hoˆpital Robert Debre´,
48 boulevard Se´rurier, 75935 Paris Cedex,
France
E-mail: [e-mail]
Tel.: [phone]
Fax: [phone]

A. Khairouni Æ K. Mazda
E. Bacheville Æ G.F. Pennec¸ot
Department of Paediatric Orthopaedics,
Hoˆpital Robert Debre´, Paris, France

J. Bloch
Department of Biostatistics,
Hoˆpital Robert Debre´, Paris,
France

S. Lamer Æ S. Dorgeret Æ A. Khairouni
K. Mazda Æ P.-Y. Brillet Æ E. Bacheville
J. Bloch Æ G.F. Pennec¸ot Æ M. Hassan
G.H. Sebag
Lariboisie`re-Saint-Louis University,
Paris, France

Abstract Background: It has been
reported that MRI using a dynamic
gadolinium-enhanced subtraction
technique can allow the early iden-
tification of ischaemia and the pat-
tern of revascularisation in Legg-
Calve´-Perthes (LCP) disease with
increased spatial and contrast reso-
lution. Therefore, dynamic gadolin-
ium-enhanced subtraction (DGS)
MRI may be a possible non-ionising
substitute for bone scintigraphy.
Objective

: The purpose of this pro-

spective study was to compare DGS
MRI and bone scintigraphy in the
assessment of femoral head perfu-
sion in LCP disease. Materials and
methods

: Twenty-six DGS MR

images and bone scintigraphies of 25
hips in 23 children were obtained at
different stages of LCP disease; three
stage I, 12 stage II, six stage III and
five stage IV (Waldenstro¨m classifi-
cation). The extent of necrosis, epi-
physeal revascularisation pathways

(lateral pillar, medial pillar, and/or
transphyseal perfusion) and meta-
physeal changes were analysed.
Results

: Total agreement between

both techniques was noted in the
depiction of epiphyseal necrosis
(kappa=1), and metaphyseal ab-
normalities (kappa=0.9). DGS MRI
demonstrated better revascularisa-
tion in the lateral (kappa=0.62) and
medial pillars (kappa=0.52). The
presence of basal transphyseal rep-
erfusion was more conspicuous with
MRI. Conclusions: DGS MRI al-
lows early detection of epiphyseal
ischaemia and accurate analysis of
the different revascularisation pat-
terns. These changes are directly re-
lated to the prognosis of LCP disease
and can aid therapeutic decision
making.

Keywords Hip Æ Legg-Calve´-
Perthes Æ MRI

proposed a scintigraphic classification system with
prognostic significance. There are two primary scinti-
graphic patterns of revascularisation. Recanalisation of
the

pre-existing

epiphyseal

vessels

allows

quicker

revascularisation and a more favourable outcome than
the neovascularisation pathway, in which new vessels
slowly grow through the physis to the femoral head
with subsequent physeal closure and deformity. In the
Conway-Dias classification, the changes seen on bone
scintigraphy precede the radiographic loss of contain-
ment of the femoral head by an average of 3 months.
The advantage of the bone scintigraphy classification
over the widely used Catterall classification, based on
the extent of involvement of the femoral head seen on
conventional radiography at the fragmentation stage, is
an earlier and higher predictive value, allowing early
and appropriate treatment [2, 5, 6].

MRI with conventional sequences has been found to

be sensitive in the diagnosis of LCP disease and in as-
sessing the extent of necrosis [7]. However, in the early
stage, MRI may fail to depict epiphyseal ischaemia [8]. It
has been reported that MRI using a dynamic gadolini-
um-enhanced subtraction technique can allow the early
identification of ischaemia and the pattern of revascu-
larisation in LCP disease with increased spatial and
contrast resolution [9]. Therefore, dynamic gadolinium-
enhanced subtraction (DGS) MRI may be a possible
non-ionising substitute for bone scintigraphy. The pur-
pose of this prospective study is to compare DGS MRI
and bone scintigraphy in the assessment of femoral head
perfusion in LCP disease.

Materials and methods

Twenty-one consecutive children with unilateral and two with bi-
lateral LCP disease were evaluated with bone scintigraphy and
MRI. The diagnosis of LCP disease was made on plain films when
typical radiographic signs were present or on bone scintigraphy in
cases of normal plain films in the initial phase. The time interval
between the two examinations was 1 month (average 16 days).
Investigations were repeated in one child. A total of 26 involved
hips (15 left, 11 right) were studied. There were 18 boys and 5 girls
(mean age 7 years, range 4–14 years), presenting with hip pain and/
or limping. According to the modified Waldenstro¨m radiographic
classification [10], 3 hips were stage I (with normal appearance of
the femoral head), 12 hips were stage II (smaller and denser femoral
head), 6 hips were at stage III (fragmentation stage) and 5 hips were
at stage IV (residual stage). The 21 normal contralateral hips were
used as a control group.

Methods

MR evaluation

MRI was performed on a 0.5-T unit (Gyrex, Elscint, Haifa, Israel)
using either a head or body coil. Imaging parameters were: FOV
27

·27 cm, 3–5-mm sections with a 1-mm gap, 192·256 matrix and

1–3 signal acquisitions. The following sequences were obtained:
sagittal spin-echo (SE), T1-weighted (T1-W; TR/TE, 500/20) and

coronal SE, proton-density and T2-weighted (T2-W; TR/TE, 2000/
20/80). A dynamic T1-W SE sequence (TR/TE, 200/20; acquisition
time 1 min) was obtained at five levels in the coronal plane.

Administration of a bolus of 0.01 mmol gadolinium tetra-aza-

cyclo-dodecane-tetra-acetic acid (DOTA, Dotarem, Guerbet, Au-
lnay, France) per kg body weight was injected by hand in less than
15 s, followed by a 10-ml saline flush. Images were acquired at the
same five levels as before, every minute for 5 min after the begin-
ning of the injection. The 1-, 2-, 3-, 4- and 5-min post-enhancement
images were then subtracted from the precontrast images, using the
subtraction function available as standard software on our console.

Bone scintigraphy

Patients were injected with technetium-99m-labelled hydroxym-
ethylene diphosphonate (7 MBq/kg). Three hours later, scintigrams
of the hips were acquired using a gamma camera (Sopha Medical
Vision) fitted with a pinhole collimator (diameter of aperture
3 mm); 100,000 counts/image were obtained. Magnified frontal
views of the hips were obtained with pinhole collimation at the
bone phase.

Patients

All children were initially seen by a senior paediatric orthopaedist.
Imaging features were retrospectively evaluated by mutual agree-
ment by two paediatric radiologists and two paediatric orthopae-
dists. According to the Herring pillar classification, the femoral
head was divided into three sectors or pillars, perpendicular to the
physis [11]. In this classification system, the pillars are derived by
noting the lines of demarcation between the central sequestrum and
the remainder of the epiphysis.

The scintigraphic activity was considered present or absent in

the lateral, central and medial pillars. The femoral neck activity was
described as homogeneous or heterogeneous, normal, increased or
decreased in comparison to the opposite side. On bone scintigra-
phy, transphyseal perfusion was characterised by widening of the
growth plate activity, giving a base filling or mushrooming ap-
pearance within the femoral head. The same criteria were applied to
MR imaging in the coronal plane. Enhancement intensity was
compared to the contralateral side (normal, absent, increased or
decreased). Transphyseal perfusion was characterised by enhance-
ment through the normally avascular physis. For each of the cri-
teria, the Kappa coefficient was estimated (with 95% confidence
interval) to assess the concordance between MRI and scintigraphy.
The Mac Nemar statistic was used to test the symmetry of each
table.

Results

Normal hips

In the 21 normal hips, gadolinium injection resulted in
early rapid (0–2 min) and intense enhancement of the
femoral head and neck and of the acetabulum, which
decreased slowly from 2 to 5 min. Enhancement was
most intense at the periphery of the femoral head and
along the femoral physis, resulting in rim-like enhance-
ment of the femoral head and linear enhancement along
the physis (Fig. 1). The central zone of the physis
remained avascular. There was total agreement between
DGS MRI and bone scintigraphy (kappa=1).

581

Abnormal hips

Necrosis

Complete agreement (kappa=1) was observed in de-
piction of an avascular zone in the 26 abnormal hips.
Avascularity was total in five cases (three stage I, two
stage II; Fig. 1) and partial in 21 cases (ten stage II, six
stage III, five stages IV). Interestingly, in the three cases
with normal plain films and normal epiphyseal marrow,
dynamic gadolinium-enhanced T1-W images without
subtraction were not able to depict the areas of avas-
cularity.

Revascularisation

Revascularisation was demonstrated in 21 hips (ten
stage II, six stage III, five stage IV). The revascularised
area exhibited increased signal intensity on the T2-W

sequence. On subtracted images, reperfusion was always
characterised by early (2 min) increased and persisting
enhancement of the revascularised area compared to the
contralateral normal hip. This enhancement pattern was
very conspicuous on subtracted images. In the depiction
of reperfusion, there were no false negatives on DGS
MRI compared to bone scintigraphy. Concordance was
poor for visualisation of the lateral pillar (kappa=0.62;
95% CI: 0.304, 0.937; Figs. 2, 3; Table 1). Compared to
dynamic MRI, bone scintigraphy did not show reper-
fusion of the lateral pillar in four cases. In only one case
did the lateral pillar have a normal height, while in three
cases height was less than 50% compared to the con-
tralateral pillar. Concordance was also poor for the
medial pillar (kappa=0.52; 95% CI: 0.222, 0.815;
Table 2). In six cases, increased vascularity was noted in
the medial pillar only on MRI. All these cases demon-
strated increased enhancement in the lateral pillar on
both techniques. Concordance between both techniques
was satisfactory for metaphyseal perfusion (kappa=0.9;
95% CI: 0.684, 1.098). Enhancement was homogeneous
in 15 cases. It was heterogeneous in ten cases with
transphyseal perfusion observed in seven of these cases
(Fig. 3). MRI localized areas of metaphyseal necrosis
more precisely in the sagittal plane. Transphyseal en-
hancement was only demonstrated by MRI, probably
because of the lack of a lateral scintigraphic view and
because this area was obscured by intense activity in the
adjacent metaphysis.

Discussion

In Legg-Calve´-Perthes disease, it is essential to identify
children who will need active treatment before femoral
head deformity occurs. Acetabular-femoral epiphysis
incongruity leads to long-term arthritis. Therefore, early
prognostic criteria are necessary for accurate treatment.
Most of the radiographic classifications are retrospective
and applicable after the primary symptoms have re-
solved, usually at the end of the fragmentation phase
[12].

The role of MRI in the diagnosis and management

of children with LCP disease is still controversial [13].
MRI is very sensitive and allows early detection when
plain films are still normal [14]. Studies comparing
MRI and bone scintigraphy show equal [15], higher
[16,17] or lesser [18] sensitivity for the detection of
LCP disease. Bone scintigraphy with pinhole collima-
tion is considered the gold-standard method for de-
tecting ischaemia of the femoral head [19]. The
localisation of radiopharmaceuticals is based on the
perfusion and metabolism of bone. Magnification
techniques must be used to identify the various scin-
tigraphic patterns that occur during the various stages
of LCP disease. Very encouraging preliminary results

Fig. 1a, b LCP disease of the left hip in the initial phase.
Radiographs were normal. The femoral head showed homogeneous
fatty high signal on T1-W images. Complete absence of perfusion
was only shown on the DGS MR image (a) and on bone
scintigraphy (b). Rim-like epiphyseal and thin linear physeal
enhancement on the MR image of the normal contralateral hip is
shown for comparison

582

were demonstrated with DGS MRI, allowing depiction
of bone perfusion with increased temporal, spatial and
contrast resolution [9].

False-negative results from conventional MRI are

reported in the initial phase of LCP disease. Mummified
bone marrow may not show any signal abnormality, and

Fig. 3a, b LCP disease of the right hip (stage III). Activity in the
medial pillar, well shown on the DGS MRI sequence (a), was
missed on bone scintigraphy (b). Revascularisation of the lateral
pillar is, however, well shown by both methods. There is
heterogeneous increased enhancement of the metaphysis by both
techniques

Fig. 2a–c Bilateral LCP disease. a Radiography shows stage II
disease in the left hip with a subchondral fracture and stage III
fragmentation on the right. There is complete agreement between
DGS MRI (b) and scintigraphy (c), which show increased
revascularisation in the lateral pillar on the left and in the medial
pillar on both sides. There is a normal appearance of the physis
with respect of the thin avascular zone on the left (typical
appearance of a favourable evolution on the left side). Basal
revascularisation on the right side is indicative of a lengthy
evolution

Table 1. Evaluation of lateral pillar revascularisation. Kappa: 0.62

MRI

+

–

Bone

+

17

0

Scintigraphy

–

4

5

Table 2 Evaluation of medial pillar revascularisation. Kappa: 0.52

MRI

+

–

Bone

+

14

0

Scintigraphy

–

6

6

583

early marrow necrosis still presents fat-like hyperinten-
sity on T1-W images [20, 21]. We noted that conven-
tional T1-W, even with fat-suppressed post-enhanced
images, usually obtained during the late vascular phase
(between 4 and 9 min after infusion) were less infor-
mative, the contrast agent already being diffusely dis-
tributed in the tissues. With the dynamic technique,
maximal marrow enhancement is depicted at approxi-
mately 2 min, allowing optimal depiction of ischaemia
[22]. Defects of vascularisation were always depicted by
comparison with the unaffected side. In the normal hip,
in addition to the rim-like enhancement of the epiphy-
seal cartilage, enhancement of the epiphyseal marrow
was depicted at 2 min in all children, even in the older
14 year old. This is noteworthy since femoral head en-
hancement is not achievable with standard MR tech-
niques in children beyond 2 years of age [23].

An important prognostic factor is the balance be-

tween the necrotic process and revascularisation, espe-
cially in the lateral pillar. MRI depicts the extent of
necrosis more clearly than pinhole scintigraphy and
earlier than changes seen on plain films [7, 8, 21, 24]. The
advantage of DGS MRI is to show the revascularisation
patterns clearly that are thought to be directly related to
prognosis [4]. This was first described by Tsao et al. [3]
on serial bone scintigraphy. Two principal mechanisms
of revascularisation may occur in LCP disease. The early
appearance of a lateral pillar is indicative of uncompli-
cated revascularisation of the femoral head. The lateral
pillar plays a key role, both through its distinctive pat-
tern of revascularisation and its mechanical keystone
property [3,11]. Sparing of the lateral pillar owing to
early recanalisation of the pre-existing epiphyseal vessels
is associated with a good prognosis. The second mech-
anism presents a slower rate of revascularisation and
healing. Involvement and collapse of the lateral pillar
secondary to extensive necrosis and late transphyseal
revascularisation result in deformity and loss of con-
tainment associated with a bad prognosis. Scintigraphic
activity extends centrally from the base with absence of a
lateral column pattern. New vessels coming from the
metaphyseal side and disrupting the normal architecture
of the growth cartilage can lead to early physeal closure
[25]. There is a third pathway, named the regression
process, corresponding to interrupted recanalisation,
because of any complication and change to a neovas-
cular pathway. Subsequently, the percentage of lateral
pillar involvement should be evaluated prospectively in
the early evolutionary period in order to allow appro-
priate management.

With DGS MRI, we were able to demonstrate rep-

erfusion patterns similar to bone scintigraphy. DGS

MRI depiction of reperfusion was very conspicuous
because of early increased and persisting enhancement
within the revascularised zones compared to normal hip
enhancement. This increased gadolinium uptake may be
due to greater vascularity, vasodilatation, increased
capillary permeability and diffusion in the reparative
process. All the scintigraphic uptakes were demonstrated
on MRI. Discrepancies corresponded to additional in-
formation provided by dynamic MRI as a result of
better spatial resolution and multiplanar slices. One can
also postulate that MRI may possibly depict revascu-
larisation within a metabolically inactive osseous area.
Medial pillar uptake was less frequently depicted on
scintigraphy, probably because the medial column can
be obscured by adjacent uptake in the acetabulum. The
lateral pillars that were not detected corresponded to a
partially collapsed lateral pillar.

DGS MRI allowed accurate visualisation of trans-

physeal revascularisation. This basal pattern of reper-
fusion was more often depicted in the anterior area,
which is known to be the site of subchondral fracture
and the site of greater vascular compromis. Transphy-
seal perfusion seems to be a predictor of growth arrest
[25].

Both techniques showed similar enhancement of the

metaphyseal region. Large metaphyseal cysts were seen
on plain radiographs, with enhancement similar to the
physeal cartilage, and thus excluding true cystic lesions.
Focal ischaemia to the ossified portion of the physeal
cartilage can leave residual unossified areas of cartilage
extending into the metaphysis. The risk for developing
growth arrest seems to be greater when metaphyseal
lesions are detected on plain radiographs [25].

When should dynamic MRI be performed? In our

experience, except for early diagnosis, MRI should be
performed between 4 and 6 months after the presenting
symptoms to obtain accurate analysis of the revascular-
isation process. Indeed, complications leading to vascu-
lar pathway changes can occur during this time interval,
so MRI should not be performed too early. The optimal
timing of dynamic MRI is not yet clearly resolved, but
this should be answered by study of a greater number of
cases. Despite its financial cost and availability, MRI
presents many advantages compared to bone scintigra-
phy. It allows accurate analysis of the femoral head de-
formity, of alteration of the cartilage, and of epiphyseal-
acetabular containment [26]. With an additional dynamic
contrast-subtracted sequence, a simple technique avail-
able on standard MR units, MRI demonstrates altera-
tion of femoral head perfusion and mechanisms of
revascularisation. All these elements are important in
dictating non-invasive appropriate management.

584

References

1. Sanchis M, Zahir A, Freeman MA

(1973) The experimental simulation of
Perthes disease by consecutive inter-
ruptions of the blood supply to the
capital femoral epiphysis in the puppy.
J Bone Joint Surg Am 55:335–342

2. Stulberg SD, Cooperman DR,

Wallensten R (1981) The natural his-
tory of LeggCalve´-Perthes disease.
J Bone Joint Surg Am 63:1095–1108

3. Tsao AK, Dias LS, Conway JJ, et al

(1997) The prognostic value and
significance of bone scintigraphy in
LCP. J Pediatr Orthop 17:230–239

4. Conway JJ (1993) A scintigraphic

classification of Legg-Calve´-Perthes
disease. Semin Nucl Med 23:274–295

5. Catterall A (1971) The natural history

of Perthes disease. J Bone Joint Surg Br
53:37–53

6. Catterall A (1981) Legg-Calve´-Perthes

syndrome. Clin Orthop 158:41–52

7. Bos CF, Bloem JL, Bloem RM (1991)

Sequential magnetic resonance imaging
in Perthes’ disease. J Bone Joint Surg
Br 73:219–224

8. Uno A, Hattori T, Noritake K, et al

(1995) LCP disease in the evolutionary
period comparison of magnetic reso-
nance imaging with bone scintigraphy.
J Pediatr Orthop 15:362–367

9. Sebag G, Ducou Le Pointe H, Klein I,

et al (1997) Dynamic gadolinium-en-
hancement subtraction MR imaging –
a simple technique for early diagnosis
of Legg-Calve´-Perthes disease: prelim-
inary results. Pediatr Radiol 27:216–
220

10. Waldenstro¨m H (1938) The first stages

of coxa plana. J Bone Joint Surg 20:559

11. Herring J, Neustadt JB, Williams JJ,

et al (1992) The lateral pillar classifica-
tion of LCP disease. J Pediatr Orthop
12:143–150

12. Van Dam BE, Crider RJ, Noyes JD,

et al (1981) Determination of the
Catterall classification in Legg-Calve´-
Perthes disease. J Bone Joint Surg Am
63:906–914

13. Henderson RC, Renner JB, Sturdivant

MC, et al (1990) Evaluation of magnetic
resonance imaging in Legg-Calve´-Per-
thes disease: a prospective, blinded
study. J Pediatr Orthop 10:289–297

14. Ranner G, Ebner F, Fotter R, et al

(1989) Magnetic resonance imaging in
children with acute hip pain. Pediatr
Radiol 20:67–71

15. Scoles PV, Yoo YS, Makley JT, et al

(1984) Nuclear magnetic resonance
imaging in Legg-Calve´-Perthes disease.
J Bone Joint Surg Am 66:1357–1363

16. Pinto MR, Peterson HA, Berquist TH

(1989) Magnetic resonance imaging in
early diagnosis of Legg-Calve´-Perthes
disease. J Pediatr Orthop 9:19–22

17. Toby EB, Koman LA, Bechtold RE

(1985) Magnetic resonance imaging of
pediatric hip disease. J Pediatr Orthop
5:665–671

18. Brody AS, Strong M, Babikian G, et al

(1991) Avascular necrosis: early MR
imaging and histologic finding in a
canine model. AJR 157:341–345

19. Conway JJ (1995) Radionuclide evalu-

ation of Legg-Calve´-Perthes disease.
In: Treves ST (ed) Pediatric nuclear
medicine, 2nd edn. Springer, Berlin
Heidelberg New York, pp 302–315

20. Elsig JP, Exner GU, Von Schulthness

GK, et al (1989) False-negative
magnetic resonance imaging in early
stages of Legg-Calve´-Perthes disease.
J Pediatr Orthop 9:231–235

21. Lahdes-Vasama T, Lamminen A,

Merikanto J, et al (1997) The value of
MRI in early Perthes’ disease: an MRI
study with a 2-year follow-up. Pediatr
Radiol 27:517–522

22. Theron J (1980) Angiography in Legg-

Calve´-Perthes disease. Radiology
135:81–92

23. Dwek JR, Shapiro F, Laor T, et al

(1997) Normal gadolinium-enhanced
MR images of the developing appen-
dicular skeleton. 2. Epiphyseal and
metaphyseal marrow. AJR 169:191–
196

24. Ducou Le Pointe H, Haddad S,

Silberman B, et al (1994) Legg-Calve´-
Perthes disease: staging by MRI using
gadolinium. Pediatr Radiol 24:88–91

25. Jaramillo D, Kasser J, Villegas-Medina

O, et al (1995) Cartilaginous abnor-
malities and growth disturbances in
Legg-Calve´-Perthes disease: evaluation
with MR imaging. Radiology 197:767–
773

26. Sebag G (1998) Disorders of the hip.

Magn Reson Imaging Clin North Am
6:627–641

585