Added Value of Diffusion-Weighted Magnetic Resonance Imaging in Differentiating Musculoskeletal Tumors Using Sensitivity and Specificity: A Retrospective Study and Review of Literature

Background: Diffusion-weighted imaging (DWI) provides added value to conventional MRI imaging in diagnosing and differentiating various benign and malignant musculoskeletal tumors. Objective: The study aims to evaluate the diagnostic efficacies of diffusion-weighted imaging along with the conventional MRI sequences for differentiating benign and malignant musculoskeletal tumors using sensitivity and specificity. Materials and methods: This retrospective study was carried out on 73 histopathologically proven patients of various musculoskeletal tumors who presented to a tertiary care center between March 2017 to October 2018. Relevant clinical examinations and MRI scan of the requested body part of the musculoskeletal system were performed. Mean apparent diffusion coefficient (ADC) values were calculated in the bone as well as soft tissue tumors after placing uniform-sized region of interest (ROI) in the non-necrotic portion of the tumor. Statistical analysis: Independent t-test and one-way analysis of variance (ANOVA) test were used to compare the mean ADC values of the various tumors with the histopathology. Receiver operating characteristic (ROC) curve analysis was done to determine the cut-off mean ADC values in the various bone and soft tissue tumors. Results: Of 73 patients with musculoskeletal tumors (benign=20, malignant = 53), 47 patients were bone tumors (benign=12, malignant=35) and 26 patients were soft tissue tumors (benign=eight, malignant=18). Mean ADC value of benign bone tumor was 1.257±0.327[SD] x 10-3mm2/s and malignant was 0.951 ± 0.177[SD] x 10-3mm2/s. The mean ADC value of benign soft tissue tumor was 1.603±0.444[SD] x 10-3mm2/s and malignant was 1.036 ± 0.186[SD] x 10-3mm2/s. The cut-off mean ADC value was 1.058 x 10-3mm2/s for differentiating benign from malignant bone tumor with a sensitivity of 83.3%, specificity of 66.7% and accuracy of 78.7% while the cut-off mean ADC value of 1.198 x 10-3mm2/s for differentiating benign from malignant soft tissue tumors with a sensitivity of 83.3%, specificity of 87.5% and accuracy of 84.6%. Conclusions: DWI with ADC mapping can be used as an additional reliable tool along with conventional MRI sequences in discriminating benign and malignant musculoskeletal tumors.


Introduction
MRI is one of the imaging modalities of choice to detect intramedullary bony abnormality even with a negative bone scan [1]. Because of excellent soft-tissue contrast and multiplanar imaging capability, the conventional MRI sequences can delineate tumor size, margins, locations, tumor necrosis, and neurovascular bundle involvement, tumoral heterogeneity and adjacent joint involvement [2]. However, adding other MRI sequences like diffusion-weighted imaging (DWI), dynamic contrast-enhanced MRI (DCE-MRI), and MR spectroscopy helps to more accurately differentiate benign from malignant bone as well as soft tissue tumors [3][4].
DWI is a functional MRI technique based on the Brownian movement of water molecules where the apparent diffusion coefficient (ADC) quantifies the Brownian movement. The presence of cellular membranes limits the diffusion in a highly cellular microenvironment, resulting in low ADC value while free diffusion takes 1, 2 3 2 4 1 place in an acellular region resulting in high ADC value [5]. Because of this property, DWI is able to provide both qualitative and quantitative assessments of intra-tumoral cellularity [6]. Various literature have shown the added advantage of DWI and ADC mapping over the conventional MRI sequences in differentiating various musculoskeletal tumors, diffuse bone marrow infiltrative lesions, benign and pathological vertebral collapse [6][7][8]. DWI with ADC mapping is also utilized for assessment of the therapeutic response after treatment of various musculoskeletal or diffuse marrow infiltrative lesions [6][7][8]. Tumor response to radiotherapy or chemotherapy showed higher ADC values as compared with the pre-treatment ADC value [9]. DWI can be a reliable additional tool over conventional MR findings in differentiating different types of bone as well as soft tissue tumors [3,[10][11]. In certain situations, conventional MRI fails to differentiate benign from malignant musculoskeletal tumors, where quantitative DWI, DCE-MRI, and MR spectroscopy play an important role [2][3][12][13]. Because of the overlapping of ADC values in the benign and malignant musculoskeletal tumors, tumor differentiation is complicated in a few situations as ADC values are affected by tumor cellularities and extracellular substances [14]. Previous studies concluded that DWI can be used as an adjunct imaging to conventional MRI sequences for characterization and differentiation of various musculoskeletal tumors [14][15].
The study aims to evaluate the diagnostic efficacies of diffusion-weighted imaging along with the conventional MRI sequences for differentiating benign and malignant musculoskeletal tumors using sensitivity and specificity.

Test methods
All patients underwent an MRI examination using a 1.5 T MR scanner (Magnatom Avanto; Siemens Medical Systems, Erlangen, Germany). An appropriate body or surface coil was used, depending upon the location and size of the lesion.
Various MRI sequences are shown in Table 1.

Analysis
Two radiologists retrospectively reviewed the MR images. During MR image analysis, the two radiologists were blinded to the clinical history or previous radiological reports of the studied patients. The size and ADC values were measured independently by the two radiologists and the mean values were used for the results.

Analysis of Conventional MR Images
We evaluated the following characteristics of a lesion like tumor sizes, margins, locations, neurovascular bundle involvement, peri-tumoral edema, tumor heterogeneity, and tumor necrosis. Tumor size was obtained from the largest dimension of a tumor. Margins of tumor classified into well defined, partially illdefined, and ill-defined. A "well defined" margin was considered when the margin of a tumor was differentiated from surrounding structures regardless of peritumoral edema. A "partially ill-defined" was classified when the margin of a tumor was mostly well defined. The location of a tumor was classified as superficial when involving skin and subcutaneous tissue or deep when the tumor was located deeper in the deep fascia. The presence of bone involvement was confirmed when there was bony cortical erosion or medullary canal involvement. Peri-tumoral edema was defined where there were peri-tumoral bright T2 weighted images (T2WI) or proton density fat-suppressed (PDFS) hyperintensities. Tumor heterogeneity was defined with mixed-signal intensities on T1WI or T2W images. Tumor necrosis was defined as an area of T2WI or short tau inversion recovery (STIR) hyperintensities that were not enhanced on post-contrast images.

ADC Calculation Analysis
ADC values were generated on a pixel by pixel basis. Minimum, maximum and mean ADC values were calculated from placing either round or elliptical ROIs, however, mean ADC values were selected for statistical analyses. ADC values were expressed in 10-3 x mm2/second. We measured the ADC value in the operating system console using multiple uniform sizes (area, minimum 10 mm2, maximum 50 mm2) at least six ROIs, where three ROIs were placed in the central non-necrotic portion of the tumor and another three in the peripheral portion. Usually, the ROI was selectively placed in the solid, enhancing, non-necrotic, and/or DWI restricted regions of a tumor. The ROI position was always checked with reference to the conventional MRI images to avoid contamination from adjacent normal-appearing bone or soft tissue. The areas of artifacts, image distortions, partial volume effect, and most peripheral margin of a tumor were avoided for ROI placement. In patients with multiple bony or soft tissue tumors, the largest lesion was selected for calculation of the mean ADC value.

Histopathology
Histopathological diagnosis of musculoskeletal tumors was established on surgically resected specimens in 29 patients, core needle biopsy specimens in 41 patients, and fine-needle aspiration (FNAC) specimens in three patients. Ultrasound-guided (USG) or CT-guided core needle biopsy specimens were obtained by using a 16-18 gauge core biopsy needle (BARD Biopsy System; Tempe, Arizona, USA). MRI studies were always performed before the biopsy or aspiration procedure. Biopsy or aspiration procedures were done from the enhancing, non-necrotic, and diffusion-restricted portion of a tumor.

Statistical Analysis
All statistical analysis was performed using Statistical Package for Social Sciences (SPSS) version 16 (IBM Corp., Armonk, NY, USA). The clinical data and different parameters of conventional MR imaging in bone and soft tissue tumors were evaluated with a chi-square test. The strength of association between the mean ADC values with the nature of tumor on histopathology was assessed using an independent t-test and oneway analysis of variance (ANOVA) test. Optimal cut-off mean ADC values of various musculoskeletal tumors were obtained from receiver operating characteristic (ROC) curve analysis.

Results
Seventy-three histologically proven patients of musculoskeletal tumors were included in this hospital-based retrospective study ( Figure 1).

FIGURE 1: Flow chart of study design
The study sample comprised 73 patients of musculoskeletal tumors (bone=47, soft tissue=26). The various histopathological types of bone and soft tissue tumors are shown in Table 2.

TABLE 2: Types of bone and soft tissue tumors in 73 patients
The anatomical locations of the tumors are shown in Table 3.

Nature of tumor Anatomical location Benign Malignant
Bone tumor Lower extremity 9 20 Upper extremity 1 7 Buttock and pelvic area 1 7 Back and chest wall -1 Soft tissue tumor Lower extremity 2 8 Upper extremity 1 9 Buttock and pelvic area 2 0 Back and chest wall 3 1

Results of clinical data and conventional MR images
Patient age, tumor margin, tumor necrosis, and adjacent joint involvement were found to be significantly related to the ability to differentiate benign and malignant bone tumors, but gender, neurovascular bundle involvement, peri-tumoral edema, and tumor heterogeneity were not ( Table 4).     Figure  2), 11 patients giant cell tumors (GCT) (Figure 3), 10 patients soft tissue sarcoma, five patients each had osteosarcoma (Figure 4), fibrosarcoma ( Figure 5), malignant fibrous histiocytoma ( Figure 6) and schwannoma, four patients had Ewings sarcoma, three patients had liposarcoma, two patients each had lymphoma (Figure 7), chondrosarcoma and fibromatosis, and one patient each had haemangioendothelioma, chondromyxoid fibroma (Figure 8), plasmocytoma, malignant haemangiopericytoma and fibroma.

Results of quantitative diffusion-weighted imaging
The

ROC curve analysis of ADC mapping
Receiver operating characteristic (ROC) curve analysis showed cut-off mean ADC value of 1.058 x 10-3mm2/s for differentiating benign from malignant bone tumor with a sensitivity of 83.3%, specificity of 66.7%, positive predictive value of 87.8%, negative predictive value of 57.1% and a diagnostic accuracy of 78.7% ( Figure 10). However two patients (7.7%) of soft tissue sarcoma showed ADC value higher than the cut-off ADC value of 1.198 x 10 -3 mm2/s while another one patient (3.8%) of schwannoma showed ADC value lower than cut-off ADC value.

Discussion
Quantitative diffusion-weighted imaging (DWI) and ADC mapping had a valuable role in characterizing and differentiating various bone and soft tissue tumors, which may improve the diagnostic accuracy in addition to the conventional MR imaging [14]. The non-contrast MRI techniques like DWI and MR Spectroscopy can be used in situations where intravenous contrast media is contraindicated. Usually, malignant tumors have low ADC values and benign tumors have high ADC values with some exceptions like giant cell tumor (GCT) and osteoblastoma, which show low ADC values [14]. The low ADC values in GCTs are due to the reduction of extracellular space from histiocytes, multi-nucleated giant cells, hemosiderin granules, and collagenous strands [3,[16][17]. High ADC values were observed in chondroid lesions, probably due to high free extracellular water content, regardless of their cellularity and histological grading [18][19][20]. The malignant chondroid tumors usually show higher ADC values than benign chondroid tumors [17] because of the high content of the chondroid matrix [16][17]. Ewing sarcoma shows the lowest ADC value in the group of sarcomas [21].
Quantitative ADC mapping helps in differentiating residual or recurrent tumor in post-treated (postirradiated) or post-operated musculoskeletal tumors [22]. Post-treated tumors with areas of higher ADC value than the pretreatment value indicate tumor cell necrosis which suggests a positive response to the therapy [6,22]. Some previous studies showed that post-treatment increasing ADC value in primary bone sarcoma correlated well with successful treatment [7,23].
In our study, conventional MR imaging parameters like tumor margin, tumor necrosis, and adjacent joint involvement were found to be significantly related to the ability to differentiate benign and malignant bone tumors, but neurovascular bundle involvement, peritumoral edema, and tumor heterogeneity were not. Neubauer et al. [30] observed the mean ADC value below 1.03 x 10-3mm2/s was a strong indicator for pediatric musculoskeletal malignancy after using a b-value of 50 and 800 s/mm2. In our study sample, we found mean ADC value below 1.058 x 10-3mm2/s is a good indicator of malignant bone tumor with a sensitivity of 83.3%, specificity of 66.7%, and diagnostic accuracy of 78.7%. So our findings are well correlated with this previous study.

Limitation
Exclusion of various benign bone as well as soft tumors in our study sample and inclusion of most of giant cell tumor of bone limited the mean ADC value results. Vast heterogeneity of the musculoskeletal tumors is a major weakness for DWI imaging and ADC mapping for diagnosis and differentiation of various musculoskeletal tumors. Therefore, a larger study sample size to confirm these findings is warranted in the future.

Conclusions
Quantitative diffusion-weighted imaging and ADC mapping helps in the evaluation of musculoskeletal tumors in conjunction with conventional MRI sequences. In our study, we found a statistically significant difference of mean ADC value between the benign and malignant bone as well as soft tissue tumors. Even though DWI and ADC mapping alone may not be useful for differentiating various benign and malignant musculoskeletal tumors because of overlapping ADC values. However, mean ADC values may serve as an additional tool while combining with the conventional MRI characteristics to diagnose and differentiate various musculoskeletal tumors.