Eur Radiol DOI 10.1007/s00330-013-2913-z
MUSCULOSKELETAL
Dynamic contrast-enhanced MR imaging for differentiation between enchondroma and chondrosarcoma T. De Coninck & L. Jans & G. Sys & W. Huysse & T. Verstraeten & R. Forsyth & B. Poffyn & K. Verstraete
Received: 13 January 2013 / Revised: 26 April 2013 / Accepted: 26 April 2013 # European Society of Radiology 2013
Abstract Objectives To determine whether dynamic contrast-enhanced (DCE) magnetic resonance imaging (MRI) can differentiate benign from malignant cartilage tumours compared to standard MRI. To investigate whether a cutoff value could be determined to differentiate enchondroma from low-grade chondrosarcoma (CS) more accurately. Methods One hundred six patients were included in this retrospective study: 75 with enchondromas (mean age=41 years) and 31 with CS (mean age=47 years). Within this population, a subgroup of patients was selected with the tumour arising in a long bone. At the time of diagnosis, the tumours were evaluated on MRI, including standard MRI, DCE-MRI, and region-of-interest (ROI) analysis to obtain information on tumour vascularisation and perfusion. Results The main cutoff value to differentiate enchondroma from CS contained a two-fold more relative enhancement compared with muscle, combined with a 4.5 (= 76°) slope value, with 100 % sensitivity and 63.3 % specificity. The prediction of CS diagnosis with DCE-MRI had 93.4 %
accuracy. The accuracy of the standard MRI parameters was equal to the DCE-MRI parameters. Conclusions Standard MRI and DCE-MRI both play an important and complementary role in differentiating enchondroma from low-grade CS. A combination of both imaging techniques leads to the highest diagnostic accuracy for differentiating cartilaginous tumours. Key Points • DCE-MRI plays an important role in differentiating benign from malignant cartilage tumours. • Retrospective study defined a threshold for 100 % detection of chondrosarcoma with DCE-MRI. • The threshold values were relative enhancement=2 and slope=4.5. • One hundred per cent chondrosarcoma detection corresponds with 36.7 % false-positive diagnosis of enchondroma. • Standard MRI is complementary to DCE-MRI in differentiating cartilaginous tumours. Keywords DCE-MRI . Dynamic . Differentiation . Chondrosarcoma . Enchondroma
Electronic supplementary material The online version of this article (doi:10.1007/s00330-013-2913-z) contains supplementary material, which is available to authorized users. T. De Coninck (*) : L. Jans : W. Huysse : T. Verstraeten : K. Verstraete Department of Radiology, Ghent University Hospital, De Pintelaan 185, 9000, Ghent, Belgium e-mail:
[email protected] G. Sys : B. Poffyn Department of Orthopaedic Surgery and Traumatology, Ghent University Hospital, De Pintelaan 185, 9000, Ghent, Belgium R. Forsyth Department of Pathology, Pathlicon Histopathology and Molecular Laboratories, Reibroekstraat 13, 9940, Evergem (Ghent), Belgium
Abbreviations and acronyms DCEdynamic contrast-enhanced magnetic resonance MRI imaging CS chondrosarcoma ROI region of interest SE spin echo GR gradient echo SI signal intensity TIC time-intensity curve AUC area under the curve ROC receiver-operating characteristic
Eur Radiol
Introduction Benign and malignant cartilaginous tumours are notoriously difficult to differentiate and represent a clinical, radiological and histological challenge. In particular lowgrade chondrosarcoma (CS) is very difficult to differentiate from benign enchondroma, at both histology and imaging [1–3]. The distinction is problematic as clinical and imaging features of benign and malignant cartilage tumours may overlap [1, 4]. Similarly, accurate prediction of the grade of CS before surgical removal could affect the type of surgery and need for adjuvant therapy, but it is not always possible with the current imaging methods. Dynamic contrast-enhanced magnetic resonance imaging (DCE-MRI) is a well-established technique that measures changes in the concentration of gadolinium in a tissue immediately after injection, and provides vascularisation and perfusion data [5, 6]. When used in cartilage tumours, it provides information on tumour vascularisation, perfusion, capillary permeability and the volume of the interstitial space. In this way, DCE-MRI may be helpful for differentiating cartilage tumours. An enchondroma with low first-pass enhancement rates and a small interstitial space displays enhancement curves similar to those of muscle. These lesions do not require biopsy. Lesions with relative enhancement rates double, triple or more than muscle most frequently represent lowgrade CS [1, 4, 7]. However, a few of these lesions will be categorised as active enchondroma by the pathologists. Therefore, a biopsy is performed in all lesions with this type of enhancement on DCE-MRI [1, 4, 7, 8], as the value of DCE-MRI in differentiating enchondroma from low-grade CS has not been proven adequately. The purpose of our study was to investigate whether DCEMRI can differentiate an actively enhancing enchondroma from a low-grade CS using DCE-MRI. Our purpose was to establish a cutoff value that more accurately predicts malignancy. We hypothesised that a cutoff value could be determined for distinction between benign and malignant cartilage tumours with DCE-MRI. Furthermore we compared the accuracy of DCE-MRI with standard MRI.
Materials and methods Patient population From 1995 to 2012, 106 patients were included in this retrospective cohort study. We included 75 patients with benign enchondroma [mean=41 years (range 12–71); 32 male, 43 female] and 31 patients with malignant CS [mean=47 years (range 22–78); 13 male, 18 female]. One female patient with enchondroma had Ollier's disease.
Within these two populations, a subgroup of cartilage tumours arising in the long bones (humerus, femur, tibia and fibula) was created. This 'long bone (LB) subgroup' consisted of 67 enchondromas [mean=44 years (range 12– 71); 28 male, 39 female] and 17 CSs [mean=49 years (range 34–65); 6 male, 11 female]. Mean follow-up was 4.85 years (range 1–17 years) for patients with enchondroma and 5.5 years (range 1–9 years) for patients with CS. All patients were treated in one centre (Ghent University Hospital). Study approval was obtained from the Ghent University Hospital Ethics Committee. The inclusion criteria for enchondroma were tumours originating within bone, histological diagnosis of enchondroma based on the biopsy specimen or typical enchondroma appearance on plain radiography and MR imaging with unchanged radiographic and MRI findings, and absence of clinical symptoms during the follow-up period (mean=4.9 years; range, 1– 17 years). The diagnosis of 37 out of 75 benign enchondromas (50.7 %) was proven at histopathological examination. Exclusion criteria for the enchondroma population were osteochondroma and osteocartilaginous exostoses. The inclusion criteria for CS were tumours originating within bone and histological diagnosis of CS based on resection specimen. Cartilage tumour distribution is illustrated in Fig. 1. Thirty-two out of 75 (42.7 %) patients had an enchondroma involving the knee, and 24 out of 75 patients (32 %) involving the proximal humerus. The CS involved the knee (distal femur, proximal tibia or proximal fibula) in 13 out of 31 patients (41.9 %). In the LB subgroup, 25 out of 67 enchondromas (37.3 %) arose in the proximal humerus and 27 out of 67 (40.3 %) in the distal femur; 7 out of 17 CS (41.2 %) arose in the distal femur. Eighteen CSs were histologically classified as low grade, 10 as intermediate grade and 3 as high grade. The following histological CS subtypes were observed: 22 classical CS, 4 juxtacortical CS, 1 myxoid CS, 1 parosteal CS, 1 secondary CS, 1 clear cell CS and 1 dedifferentiated CS. Patients with CS were treated with a surgical resection or extracorporeal irradiation therapy. The resection was limb salvaging where possible. No chemotherapy was administered as CSs are resistant to chemotherapy [9]. For patients with a low-grade CS, cryosurgery combined with curettage was also a treatment option [10]. Patients with enchondroma were subject to observation and follow-up. Patients with large symptomatic lesions underwent a total resection with curettage and bone grafting. Of all 75 patients with enchondroma, none developed a secondary CS during the follow-up period. One patient (1.3 %) died of a cause unrelated to his enchondroma. Thirty out of 31 patients (97 %) presented with a localised CS. One patient (3 %) presented with lung metastases. Two out of 31 patients (6.5 %) developed metastases in the course of their disease. The three patients with metastasised
Eur Radiol Fig. 1 Skeletal distribution of cartilage tumours. The distribution is demonstrated for both the whole group, and the long bone subgroup. The distribution is mentioned in terms of percentage
disease were diagnosed with an intermediate- or high-grade CS. Eight out of 31 patients (25.8 %) with CS developed local recurrence of their CS; only 1 of the patients had an LB CS. Fifty per cent of all CS recurrences were primarily located in the pelvis. After a mean 5.5-year follow-up (range 1–9 years), 27 out of 31 patients (87.1 %) were free of disease. Four out of 31 patients (12.9 %) died of local recurrence, progressive disease and metastases. One lesion was high-grade, two were intermediate grade and one patient had a clear cell CS. Mean survival time from diagnosis of CS until death was 7.8 years. Of all 31 CS patients, not a single patient with a CS of the long bone died, the clear cell CS excluded (mean follow-up LB CS=5.3 years, range 1–9 years).
slice thickness=6 mm; flip angle 12°; FOV 160–360 mm; matrix 256×256 mm). Every 1.1 s an image was acquired; thus, a total of 120 images were made at one level through the lesion. Five minutes after gadolinium administration, sagittal and coronal T1-weighted SE images (Fig. 3a and b) were obtained using identical imaging parameters to those used for the precontrast T1-weighted sequences. Image post-processing The DCE-MRI images were post-processed on a Siemens Leonardo workstation (Siemens, Erlangen, Germany) using the mean-curve application. These images, covering the first 2 min after bolus injection during which the first pass of the
MR imaging Preoperative magnetic resonance (MR) images were obtained at initial presentation on 1.5-T MR units (Magnetom Avanto or SymphonyTim, Siemens, Erlangen, Germany). Depending on the tumour volume, a surface or body flexed array coil was used. Precontrast imaging included longitudinal (coronal or sagittal, depending on the location of the lesion) and axial T1-weighted spin-echo (SE) sequences (TR=600 ms; TE=20 ms; slice/gap: 4– 8 mm/0.4–2 mm; FOV: 200–400 mm; matrix size: 256× 256 - 512×384), followed by longitudinal and axial intermediate-weighted SE images with fat saturation (TR= 4,400 ms; TE=43 ms; slice/gap: 4–8 mm/0.4–2 mm; FOV: 200–400 mm; matrix size: 512×240–384; Fig. 2a and b). For DCE-MRI a single slice in the coronal or sagittal plane was chosen, based on the T1-weighted SE images. This plane always exhibited a representative intramedullary and, if present, extramedullary tumour component. DCEMRI was obtained by rapid intravenous bolus injection (injection rate: 5 ml/s) of 0.1 mmol/kg of body weight of Gd-DTPA (Magnevist®; Bayer Schering Pharma, Berlin, Germany), immediately followed by a 20-ml saline flush. The injection was started simultaneously with the T1weighted gradient-echo (GRE) sequence (Turbo Flash 2D sequence, Siemens; TR=1,100 ms; TE=3 ms; TI=560 ms;
Fig. 2 a Precontrast longitudinal T1-weighted spin-echo (SE) MR image of a low-grade chondrosarcoma in the diaphysis of the femur. b Longitudinal T1-weighted SE MR image 5 min after contrast medium administration. The white arrows and dotted lines delineate the CS on both a and b
Eur Radiol
contrast agent through all tissues occurred, were evaluated with the region-of-interest (ROIs) method [11]. The ROIs were selected and manually encircled with a cursor on the display screen (Fig. 4a). Whole tumour ROI assessment may be inappropriate, particularly for the evaluation of malignant lesions where heterogeneous areas of enhancement are diagnostically important [11]. Therefore, several ROI samples were drawn, including ROIs of the most enhancing tumour part and the complete tumour. Muscle and arterial tissues were plotted as reference tissues (Fig. 4a). A time-intensity curve (TIC; Fig. 4b) was then created, whereby the signal intensity (SI) of the selected region is plotted against time. These TICs are a graphical representation of
the pharmacokinetics of the contrast medium during and immediately after the first pass and represent microvascular perfusion [5]. The changes in SI of each ROI were plotted against time, reflecting the different enhancement patterns of various tissues and tumour components. From the generated TIC (Fig. 5), the following parameters were calculated with Engauge (version 4.1 © 2002, Mark Mitchell): time of onset of the first pass of the contrast medium (Tstart) with corresponding signal intensity (SIstart) and the (time to) maximal signal intensity (Tmax and SImax). Those semiquantitative parameters, depicting the wash-in, included the steepest slope of the first pass enhancement, absolute and relative maximum enhancement. They were calculated through the formulas illustrated in Fig. 6. The slope is representative of maximum tissue enhancement during first pass and is mainly determined by vascularisation and tissue perfusion [11] (Table 1). Absolute enhancement is the maximum increase in SI compared with the SI of the tissue before arrival of the contrast medium (=SIstart). To minimise the difference in absolute values between patients, the maximum SI was standardised to the maximum SI of muscle tissue, resulting in the relative enhancement. Thus, the relative enhancement compares the absolute enhancement between tissues. In our study, the enhancement of the tumour and the enhancement of the unaffected muscle were compared. Because the DCE-MR images were T1-weighted, tissue with higher fat content had a higher intrinsic SI, regardless of vascularisation. Consequently the absolute values of different tissue types could not be compared. To correct for these values, the absolute values were standardised to start
Fig. 4 a The region-of-interest (ROI) zones were selected and manually encircled with a cursor on the display screen. The red ROI represents the most enhancing part of this enchondroma. Muscle (yellow) and artery (blue) were also manually encircled and served as reference tissues. b The signal intensity of the selected ROI is plotted against time in a time-intensity curve (TIC). These
TICs are indicative of vascular ingrowth. Several ROIs are calculated: the most enhancing tumour part (red), muscle tissue (yellow) and artery (blue) were plotted as reference tissue. Changes in the signal intensity (SI) of each ROI were plotted against time, reflecting the different enhancement patterns of various tissues and tumour components
Fig. 3 a Precontrast coronal T1-weighted spin-echo (SE) MR image of an enchondroma in the proximal humerus. b Coronal T1-weighted SE MR image 5 min after contrast medium administration. The white enhancing zones are the most active zones in the enchondroma
Eur Radiol Fig. 5 A time-intensity curve displaying the time of onset of the first pass of the contrast medium (Tstart) with corresponding signal intensity (SIstart) and the (time to) maximal signal intensity (Tmax and SImax). From these parameters the steepest slope of the first-pass enhancement, the absolute and relative maximum enhancement can be calculated
SI, resulting in relative values. The conversion of the percentage slope to slope in degrees was calculated using the following formula: Slope in Degrees=Tan-1 (Slope Percent/100). Radiological outcome variables Tumour volume was measured on the coronal and axial T1weighted SE images. The maximum height, width and depth of the entire tumour were measured. Tumour volume was calculated with the formula of an elliptic mass: volume=π/6× Fig. 6 Formula for the calculation of the semiquantitative parameter 'steepest slope of the first-pass enhancement'. The slope is representative of maximum tissue enhancement during first pass and is mainly determined by vascularisation and tissue perfusion *The multiplication factor 1.1 was necessary because the temporal resolution of the DCE study was 1.1 s. SI = signal intensity, T = tumour, M = muscle, Abs Enh = absolute enhancement, t = time in seconds
height×width×depth [12]. Signal intensity was measured on the T1- and T2-weighted SE images, in which the largest cross-sectional area of the tumour was presented. The intensity was defined as hypointense, isointense or hyperintense, relative to the normal muscle tissue. Several MR-parameters derived from the literature [8, 13, 14] were used to describe the lesions at standard MRI (Table 1). Slope and relative enhancement were used as dynamic imaging criteria for microvascularisation. Slope correlates
(Siend T– Sistart T) Steepest slope = --------------------------------- x 1100/1000* in %/s (Sibaseline T x t) (Simax T) – (Sistart T) Absolute enhancement = ----------------------------- x 100 in % Sistart T Abs Enh x (Sistart M) Relative enhancement = ----------------------------- in % (Simax M) – (Sistart M)
Eur Radiol Table 1 MR features of cartilaginous tumours. Features of both enchondroma and CS, determined on standard MR images
Enchondroma
CS
P value
PPV (%)
n
%
n
%
8/65 7/65 10/65 3/65 2/65
12.3 10.7 15.4 4.6 3.1
2/14 12/14 8/14 3/14 4/14
14.3 85.7 57.1 21.4 28.6
1.000
