G Model EURR-7047; No. of Pages 7

ARTICLE IN PRESS European Journal of Radiology xxx (2015) xxx–xxx

Contents lists available at ScienceDirect

European Journal of Radiology journal homepage: www.elsevier.com/locate/ejrad

Differentiation of benign and malignant skeletal lesions with quantitative diffusion weighted MRI at 3 T Shivani Ahlawat a,∗ , Paras Khandheria a,1 , Ty K. Subhawong b,2 , Laura M. Fayad a,1 a The Russell H. Morgan Department of Radiology & Radiological Science, The Johns Hopkins University School of Medicine, 601 North Wolfe Street, Baltimore MD 21287, United States b Department of Radiology (R-109), University of Miami Leonard M. Miller Miami, FL 33101, United States

a r t i c l e

i n f o

Article history: Received 23 May 2014 Received in revised form 26 November 2014 Accepted 21 February 2015 Keywords: Diffusion-weighted imaging Bone tumors ADC map Musculoskeletal

a b s t r a c t Objectives: To investigate the accuracy of quantitative diffusion-weighted imaging with apparent diffusion coefficient (ADC) mapping for characterizing bone lesions as benign or malignant. Methods: At 3 T, 31 subjects with intramedullary lesions imaged by DWI (b-values 50, 400, 800 s/mm2 ) were included. ADC values (minimum, mean, maximum) were recorded by three observers independently. Interobserver variability and differences between ADC values in benign and malignant lesions were assessed (unpaired t-test, receiver operating characteristic (ROC) analysis). Results: Of 31 lesions, 18 were benign (osteoblastic (n = 1), chondroid (n = 6), cysts (n = 4), hemangiomatosis (n = 1), fibrous (n = 3), eosinophilic granuloma (n = 1), giant cell tumor (n = 1), osteomyelitis (n = 1)) and 13 were malignant (primary (n = 5), metastases (n = 8)). Overall, there were higher minimum (1.27 × 10−3 mm2 /s vs 0.68 × 10−3 mm2 /s, p < 0.001), mean (1.68 × 10−3 mm2 /s vs 1.13 × 10−3 mm2 /s, p < 0.001), and maximum (2.09 × 10−3 mm2 /s vs 1. 7 × 10−3 mm2 /s, p = 0.03). ADC values in benign lesions compared with those in malignancies. ROC analysis revealed areas under the curve for minimum, mean, and maximum ADC values of 0.91, 0.85, and 0.71, respectively. ADC measurements were made with high inter-observer concordance ( = 0.83–0.96). Conclusion: Quantitative ADC maps may have predictive value for the characterization of bone lesions. Benign lesions generally have higher minimum, mean, and maximum ADC values than malignancies, with the minimum value offering the highest accuracy for characterization. © 2015 Elsevier Ireland Ltd. All rights reserved.

1. Introduction Magnetic resonance imaging (MRI) plays a central role in the evaluation of bone tumors, primarily for the assessment of lesion extent. With conventional MRI techniques, which include T1-weighted (T1), fluid-sensitive and contrast-enhanced T1 sequences, exquisite contrast between a skeletal lesion and surrounding normal marrow is achieved. However, MRI techniques lack specificity with lesion characterization and other modalities (radiography) are utilized for this purpose [1,2]. DWI is a noncontrast, functional MRI technique that has been investigated for

∗ Corresponding author. Tel.: +1 443 287 6032; fax: +1 410 502 6454. E-mail addresses: [email protected] (S. Ahlawat), [email protected] (P. Khandheria), [email protected] (T.K. Subhawong), [email protected] (L.M. Fayad). 1 Tel.: +1 443 287 6032; fax: +1 410 502 6454. 2 Tel.: +1 305 585 7500.

the characterization of tumors throughout the body [3–5]. The apparent diffusion coefficient (ADC) is a quantitative measure of Brownian movement and has been established as a marker of tumor cellularity: Low ADC values reflect highly cellular microenvironment, whereas high ADC values are observed in acellular [3,6]. In the musculoskeletal system, quantitative DWI has been explored for lesion characterization, although most studies have focused on the characterization of soft tissue masses, with mixed results regarding the utility of DWI for this purpose [6–12]. With regard to skeletal lesions, prior investigations have utilized DWI in the setting of localized and whole body imaging for the detection of osseous metastases and multiple myeloma [13–27], as well as for the determination of treatment response [10–12]. There is a paucity of information on the role of DWI for characterizing skeletal lesions as benign or malignant, with limited populations and variable results [28,29]. Our hypothesis was that quantitative DWI offers information valuable for characterizing skeletal lesions for malignancy. Therefore, the purpose of this study was to investigate the accuracy of quantitative DWI with ADC mapping at 3 T for

http://dx.doi.org/10.1016/j.ejrad.2015.02.019 0720-048X/© 2015 Elsevier Ireland Ltd. All rights reserved.

Please cite this article in press as: Ahlawat S, et al. Differentiation of benign and malignant skeletal lesions with quantitative diffusion weighted MRI at 3 T. Eur J Radiol (2015), http://dx.doi.org/10.1016/j.ejrad.2015.02.019

G Model EURR-7047; No. of Pages 7

ARTICLE IN PRESS S. Ahlawat et al. / European Journal of Radiology xxx (2015) xxx–xxx

2

Fig. 1. Flow chart with inclusion and exclusion criteria.

characterizing bone lesions as benign or malignant, and observe if there was any correlation of ADC values with specific histologies. 2. Materials and methods In this IRB-approved retrospective study, the imaging of 31 subjects with intramedullary lesions who had undergone conventional MRI with T1-weighted, T2-weighted, and contrast-enhanced T1-weighted sequences as well as quantitative DWI with ADC mapping, was reviewed. The ADC values of the bone lesions and the adjacent normal bone were recorded. ADC measurements of benign and malignant bones lesions were compared, as were the ADC values of specific histologies. 2.1. Subject population Consecutive subjects who were treated in Orthopedic Oncology clinics between September 2009 and August 2013 were sought. One observer reviewed all available clinical records on the subjects (demographic data, information on clinical follow-up, pathologic results) and selected subjects with skeletal lesions (primary or metastatic) only. Inclusion criteria were subjects who had a bone lesion that was referred for de novo evaluation by MRI prior to any treatment, lesion size greater to or equal to 1 cm (in order to ensure an accurate measurement of the ADC values within a region of interest [ROI]) and subjects who had undergone MRI with a uniform DWI protocol (acquired with three b-values of 50, 400, and 800 mm2 /s). Benign lesions were diagnosed by histology (when

available) or follow-up demonstrating radiological and clinical stability of at least 6 months. Malignant lesions were diagnosed by histology or were presumed malignant in the setting of known metastatic carcinomatosis and multiple bone lesions; (the patient was included if there were more than 10 metastatic bone lesions with at least one biopsy-proven lesion). Exclusion criteria were lesions that had already been treated with chemotherapy or radiation therapy or surgery, benign or malignant lesions complicated with fracture or percutaneous biopsy (as blood products would be present and alter DWI characteristics), and lesions that were imaged without a complete DWI examination. Fig. 1 details subject selection for this study.

2.2. MRI protocol All examinations were performed at 3 T (Verio; Siemens Medical Systems, Malvern, PA, USA). T1-weighted (TR/TE 960/9, SL 5–6 mm) and fat suppressed (FS) T2-weighted sequences (TR/TE 3600–4280/70, SL 5–6 mm) in the axial and coronal planes were obtained. DWI was performed in the axial plane using a spin-echo, single-shot echo-planar imaging (EPI) sequence. An inversion-recovery pulse (inversion time [TI] = 180 ms) was used to exclude severe chemical-shift artifacts. The following parameters were used for DWI: TR = 760 ms, TE = 80 ms, NEX = 2, gradient strength = 25 mT/m, FOV = 180–250 mm2 , matrix size = 256 × 256 pixels, section thickness = 5 mm, interslice gap = 1 mm, section levels = 30; a partial Fourier transform and EPI factor = 88 was used. The b-values used were 50, 400, and 800 s/mm2 . ADC maps were calculated using a monoexponential fit with inline software

Please cite this article in press as: Ahlawat S, et al. Differentiation of benign and malignant skeletal lesions with quantitative diffusion weighted MRI at 3 T. Eur J Radiol (2015), http://dx.doi.org/10.1016/j.ejrad.2015.02.019

G Model EURR-7047; No. of Pages 7

ARTICLE IN PRESS S. Ahlawat et al. / European Journal of Radiology xxx (2015) xxx–xxx

3

from Siemens (Syngo MapIT). 3-D T1-FS sequence (volume interpolated breath-hold examination (VIBE), TR/TE 4.6/1.4, flip angle 9.5, SL 1 mm) was obtained before and after the administration of 0.1 mmol/kg body weight of gadolinium-diethylenetriamine pentaacetic acid (DTPA) (Magnevist; BayerSchering, Berlin, Germany). Static post contrast subtraction imaging was also obtained, with subtraction of the pre-contrast from post-contrast images. 2.3. Image analysis Three observers (one musculoskeletal fellowship-trained radiologist with 2 years of experience in musculoskeletal imaging, one fellowship-trained radiologist with 12 years of experience in musculoskeletal tumor imaging and one physician performing a mini-fellowship in musculoskeletal imaging) reviewed all imaging independently. First, readers recorded lesion location from the T1 images; the presence of a skeletal lesion was determined as a welldefined signal abnormality causing marrow-replacement by T1 images. Next, the quality of the DWI sequences and ADC maps was recorded, using a 1–3 scale (1 = non-diagnostic with >25% artifact on images, 2 = diagnostic with between < = 25% artifact, 3 = diagnostic with no substantial artifact). Third, each reader, independently, constructed an ROI within each lesion on the ADC map (on axial views) on up to three slices (cranial, middle and caudal) including areas of lowest ADC signal, as described by prior methodology [12]. An ROI was also placed in the adjacent normal bone, defined by correlation with T1 images. Minimum, maximum, and mean ADC values were recorded from intralesional ROIs as well as from ROIs placed in nearby normal bone, and the values across slices were averaged. Next, readers assessed the anatomic imaging characteristics of each lesion. The following features were observed on the noncontrast sequences (T1W, T2W): signal intensity (hypointense; isointense; hyperintense to muscle), as well as signal heterogeneity (homogeneous; 75% heterogeneous). Contrast enhancement was categorized as chondroid-type (with lobular enhancement), solid (with complete enhancement), peripheral (with rim enhancement), thin internal septal enhancement or lack of any enhancement. Readers also recorded the presence or absence of mineralization in each lesion by radiography or CT if it was available; this record was made because a mineralized matrix could potentially cause susceptibility artifact and impact the ADC values within a lesion. 2.4. Statistical analysis Descriptive statistics were reported. Differences in age between subjects with benign and malignant bone lesions were compared using Student’s t-test. Differences in gender between subjects with benign and malignant bone lesions were compared using Fisher’s exact test. All continuous variables (minimum, maximum, and mean lesion ADC value) were compared for benign and malignant bone lesions by Student’s t-test using means derived from the two readers’ measurements (3 observations per reader). Average concordance correlation coefficient was calculated using an average of the ADC values for reader 1 and 2 to assess the inter-rater comparison. Receiver-operating characteristic (ROCs) were constructed for minimum, maximum, and mean ADC values for differentiating benign and malignant bone lesions for each observer. Threshold ADC values with sensitivity, specificity, accuracy and positive likelihood ratios (LR+) were calculated for the minimum, mean and maximum ADC values. In addition, the ADC values of the lesions were compared with normal adjacent bone using Student paired t-test. Differences were considered significant at p < 0.05.

Fig. 2. Receiver operating characteristic (ROC) analysis of minimum, mean and maximum ADC values for distinguishing between benign and malignant bone lesions. Performance for each variable is given by area under the curve (AUC); better performance is indicated by high AUC. Mean ADC values performed the best (p = 0.02).

3. Results Of the 31 lesions, 18 were benign and 13 were malignant. Of the 18 benign lesions, 10 were biopsied with pathological confirmation. Similarly, of the 13 malignant lesions, 7 were pathologically confirmed. Table 1 lists the histologies of each lesion as well as the recorded ADC values for each lesion. There was a difference in age between subjects with benign lesions (mean age 26.4 ± 19.7 years, age range 4–65 years) and those with malignant lesions (50.8 ± 22.1 years, age range 9–74 years), p = 0.003. There was also a difference in gender composition between the two groups (61% [11/18] of benign bone tumors vs 38% [5/13] of malignant bone tumors occurred in females) but this did not reach statistical significance (p = 0.3, Fisher exact). The quality of the DWI was diagnostic with a score of 3 (diagnostic with no substantial artifact) in 100% of the cases. Table 1 shows the ADC values of each lesion, while Table 2 shows a summary of the ADC values recorded from the benign and malignant histologies in the study population, as well as the adjacent normal bone marrow. Regarding interobserver variability for the measurements, the average concordance correlation coefficient (c ) was high for minimum ADC value (c = 0.83), and very high for mean and maximum ADC values for the three readers (c = 0.95 and 0.96, respectively). Overall, there were higher minimum (1.27 × 10−3 mm2 /s vs 0.68 × 10−3 mm2 /s, p < 0.001), mean (1.68 × 10−3 mm2 /s vs 1.13 × 10−3 mm2 /s, p < 0.001), and maximum (2.09 × 10−3 mm2 /s vs 1. 7 × 10−3 mm2 /s, p = 0.03). ADC values in benign lesions compared with those in malignancies, respectively. The ADC values of normal bone adjacent to the intramedullary lesions, including minimum, mean and maximum values were recorded and compared with the ADC values of the bone lesions. All normal bone ADC values were lower compared to those of bone lesions. Notably, when age was dichotomized to ≤30 or >30, there was no significant difference between the ADC values of normal bone with respect to the subject age (p = 0.55, unpaired t-test). ROC analysis showed the minimum ADC as providing the highest accuracy, with areas under the curve for minimum, mean, and maximum ADC values of 0.91, 0.85, and 0.71, respectively (Fig. 2). Threshold analysis yielded optimal threshold values of 0.9 × 10−3 mm2 /s (minimum ADC) and 1.4 × 10−3 mm2 /s (mean ADC) as imparting sensitivities of 92% and 77% and specificities of 78% and 78% for differentiating benign and malignant histology, respectively, (as shown in Table 3). Figs. 3 and 4 are examples of benign and malignant lesions assessed by DWI with ADC mapping

Please cite this article in press as: Ahlawat S, et al. Differentiation of benign and malignant skeletal lesions with quantitative diffusion weighted MRI at 3 T. Eur J Radiol (2015), http://dx.doi.org/10.1016/j.ejrad.2015.02.019

G Model

ARTICLE IN PRESS

EURR-7047; No. of Pages 7

S. Ahlawat et al. / European Journal of Radiology xxx (2015) xxx–xxx

4

Table 1 Summary of demographics and individual bone lesion characteristics. Patient

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31

Age (years)

4 65 35 61 31 42 13 9 16 10 32 8 51 11 49 28 11 11 36 74 68 59 69 73 50 35 33 80 19 9 64

Gender (M/F)

F F M F F F F M M F M F M M F F M F F M M M M M F F M M F M F

Diagnosis

Enchondroma Enchondroma Enchondroma Enchondroma Enchondroma Chondroblastoma Unicameral bone cyst Unicameral bone cyst Aneurysmal bone cyst Aneurysmal bone cyst Hemangiomatosis of bone Infection Fibrocystic lesion Fibrous dysplasia Fibrous dysplasia Giant cell tumor Langerhans cell histiocytosis Osteoblastoma Breast metastases Prostate metastases Squamous cell carcinoma metastases RCC metastases Prostate metastases Prostate metastases RCC metastases Adenocarcinoma, unknown primary Plasmacytoma Angiosarcoma Osteosarcoma Osteoarcoma Osteosarcoma

Classification benign/malignant

Benign Benign Benign Benign Benign Benign Benign Benign Benign Benign Benign Benign Benign Benign Benign Benign Benign Benign Malignant Malignant Malignant Malignant Malignant Malignant Malignant Malignant Malignant Malignant Malignant Malignant Malignant

Location

Phalanx Femur Femur Phalanx Femur Femur Metacarpal Fibula Acetabulum Tibia Calcaneus Tibia Femur Femur Tibia Femur Ilium Sacrum Humerus Sacrum Femur Femur Ilium Femur Tibia Femur Femur Femur Humerus Humerus Ilium

Mineralization

Absent Chondroid Chondroid Absent Absent Chondroid Absent Absent Absent Absent Absent Absent Absent Absent Absent Absent Absent Osteoid Absent Sclerotic Absent Absent Sclerotic Absent Absent Absent Absent Absent Osteoid Osteoid Osteoid

ADC value (×10−3 mm2 /s)

Minimum

Mean

Maximum

1.7 1.1 1.3 1.5 1.7 0.8 1.7 0.8 1.3 1.0 1.6 1.2 1.8 1.2 1.4 0.6 0.7 0.7 0.8 0.4 0.9 0.5 0.6 0.7 0.5 0.8 0.5 0.7 0.5 1.1 0.8

1.9 1.6 1.7 1.8 2.0 1.3 2.2 1.5 1.9 2.1 2.0 1.4 2.2 1.6 1.6 1.2 1.1 1.1 0.9 0.6 1.5 1.0 0.8 1.4 1.0 1.0 0.8 1.7 0.9 1.9 1.2

2.0 1.9 2.0 1.9 2.4 1.9 2.6 2.4 2.3 2.9 2.2 1.6 2.6 2.1 1.7 2.2 1.4 1.5 1.2 0.7 1.9 1.7 0.9 1.8 2.0 1.2 1.1 2.7 1.7 2.6 2.1

Table 2 Comparison of ADC values of benign lesions, malignancies and normal marrow.

Benign bone lesions Malignant bone lesions Normal bone marrow p-Value (benign vs malignant) AUC a

Minimum ADC value (×10−3 mm2 /s)

Mean ADC value (×10−3 mm2 /s)

Maximum ADC value (×10−3 mm2 /s)

1.27 (±0.47) 0.68 (±0.21) 0.02 (±0.08)a