Abstract
Objective:
To evaluate multiparametric MRI for differentiating benign and malignant soft tissue tumors.
Methods:
This retrospective study included 67 patients (mean age, 55 years; 18–82 years) with 35 benign and 32 malignant soft tissue tumors. Intravoxel incoherent motion diffusion-weighted imaging (IVIM-DWI)-derived parameters (D, D*, f), apparent diffusion coefficient (ADC), and dynamic contrast-enhanced (DCE)-MRI parameters (Ktrans, Kep, Ve, iAUC) were calculated. Myxoid and non-myxoid soft tissue tumors were divided for subgroup analysis. The parameters were compared between benign and malignant tumors.
Results:
ADC and D were significantly lower in malignant than benign soft tissue tumors (1170 ± 488 vs 1472 ± 349 µm2/s; 1132 ± 500 vs 1415 ± 374 µm2/s; p < 0.05). Ktrans, Kep, Ve, and iAUC were significantly different between malignant and benign soft tissue tumors (0.209 ± 0.160 vs 0.092 ± 0.067 min−1; 0.737 ± 0.488 vs 0.311 ± 0.230 min−1; 0.32 ± 0.17 vs 0.44 ± 0.28; 0.23 ± 0.14 vs 0.12 ± 0.09, p < 0.05, respectively). ADC (0.752), D (0.742), and Kep (0.817) had high AUCs. Subgroup analysis showed that only Ktrans, and iAUC were significantly different in myxoid tumors, while, ADC, D, Ktrans, Kep, and iAUC were significantly different in non-myxoid tumor for differentiating benign and malignant tumors. D, Kep, and iAUC were the most significant parameters predicting malignant soft tissue tumors.
Conclusion:
Multiparametric MRI can be useful to differentiate benign and malignant soft tissue tumors using IVIM-DWI and DCE-MRI.
Advances in knowledge:
1. Pure tissue diffusion (D), transfer constant (Ktrans), rate constant (Kep), and initial area under time–signal intensity curve (iAUC) can be used to differentiate benign malignant soft tissue tumors.
2. Ktrans and iAUC enable differentiation of benign and malignant myxoid soft tissue tumors.
Introduction
It is not uncommon to encounter problems distinguishing benign from malignant soft tissue tumors in routine practice. It can be difficult to differentiate between benign and malignant soft tissue tumors with conventional MRI alone because morphologic features and signal intensities of benign and malignant soft tissue tumors may overlap.1–5 Recently, diffusion-weighted imaging (DWI) has been used to evaluate soft tissue tumors. Calculation of apparent diffusion coefficient (ADC) values for random water movement in tissues can reflect the pathologic status of diseased tissue.6–10 DWI based on an intravoxel incoherent motion (IVIM) biexponential model by Le Bihan et al11 can more accurately separate microcapillary perfusion from pure tissue diffusion (D). Perfusion characteristics (pseudodiffusion coefficient, D*) and their volume fraction (perfusion fraction, f) can be obtained simultaneously. IVIM-DWI hasbeen applied in soft tissue tumors to demonstrate the feasibility of differentiation between benign and malignant soft tissue tumors.12,13 Dynamic contrast-enhanced (DCE)-MRI has been reported as potentially helpful in characterizing soft tissue tumors.14–16 When a pharmacokinetic model by Tofts17 is applied, DCE-MRI can derive three main quantitative parameters: transfer constant (Ktrans), rate constant (Kep), and extravascular extracellular volume fraction (Ve), as well as the semi-quantitative parameter of initial area under the time–signal intensity curve (iAUC). The values of these pharmacokinetic parameters for DCE-MRI have rarely been evaluated in soft tissue tumors.18,19
Despite these developments, there have been inconsistent reports about using DWI, IVIM, or DCE-MRI for differentiating benign and malignant soft tissue tumor and few reports fully evaluating quantitative multiparametric MRI including IVIM-DWI and DCE-MRI for differentiating benign and malignant soft tissue tumors. Therefore, the purpose of our study was to determine the value of multiparametric MRI for differentiation between benign and malignant soft tissue tumors.
Methods and materials
Study population
This retrospective study was approved by our Institutional Review Board, and the requirement for informed consent was waived. From May 2014 to February 2018, a total of 130 patients was enrolled in the study with the following criteria: (i) first diagnosis of a soft tissue mass (images were examined before histologic biopsy and/or initiation of treatment), (ii) soft tissue tumors with pathologic confirmation using histologic biopsy and/or examination of the surgical specimens, and (iii) underwent 3 T MRI including IVIM-DWI and DCE-MRI. Among these 130 patients, 63 were excluded based on the following exclusion criteria: (i) 30 non-neoplastic lesions (21 ganglion cysts, 6 epidermoid inclusion cysts, and 3 metastases), (ii) 24 well-differentiated adipocytic tumors (20 lipomas and 4 well-differentiated liposarcomas), (iii) 4 soft tissue tumors less than 0.5 cm in diameter (2 fibromatosis and 2 glomus tumors), (iv) 4 with unsatisfactory image quality in most sequences due to artifacts (three tenosynovial giant cell tumors and one fibromatosis), and (v) one intermediate malignancy (dermatofibrosarcoma protuberans). Finally, the remaining 67 patients (mean age, 55 ± 15 years; range, 18–82 years) with soft tissue tumors constituted the study populationof 30 males (mean age, 54 ± 16 years; range, 18–82 years) and 37 females (mean age, 57 ± 15 years; range, 26–82 years). Part of the population of this study (n = 17) originates from a previous study 20 with different measurements and aims to determine the diagnostic value of multiparametric MRI for differentiating benign from malignant soft tissue tumors. Of these patients, six patients (angioleiomyoma, fibroma,fibromatosis,glomus tumor, tenosynovial giant cell tumor, and malignant melanoma) were further excluded from IVIM-DWI analysis due to artifacts. The DCE-MRI analysis included all 67 patients (Figure 1).
Figure 1.
Study flowchart. DCE-MRI, dynamic contrast-enhanced magnetic resonance imaging; IVIM-DWI, intravoxel incoherent motion diffusion-weighted imaging
Since it has been reported that a myxoid tumor matrix forms parts of many benign and malignant soft tissue tumors,21 we decided to compare quantitative parameters between patients with and without myxoid tumor matrix. For these reasons, the tumors were divided into 19 myxoid and 48 non-myxoid soft tissue tumors for subgroup analysis. Myxoid soft tissue tumors included myxoid liposarcomas (n = 3) and myxofibrosarcomas (n = 3), schwannomas (n = 8), myxomas (n = 2), a neurofibroma (n = 1), a low grade fibromyxoid sarcoma (n = 1), and a malignant peripheral nerve sheath tumor (n = 1).
MRI acquisition
Images were obtained using a 3 T system (MAGNETOM Verio, Siemens Healthineers, Erlangen, Germany). Field of view varies by body part. Conventional MRI was obtained with coil adjustments. Axial turbo spine echo (TSE) T1 weighted and TSE T2 weighted images with and without fat suppression, at least one fat suppressed T2 weighted longitudinal scan, and contrast-enhanced longitudinal and axial TSE T1 weighted images with fat suppression were obtained.
IVIM-DWI was obtained using single-shot spin echo echoplanar imaging. Encoding was performed in three orthogonal directions. A series of nine b-values (0, 25, 50, 75, 100, 200, 300, 500, and 800 s/mm2) was applied. The acquired imaging data were post-processed to obtain ADC, and the IVIM-derived parameters including D (diffusion coefficient), D* (pseudodiffusion), and f (perfusion fraction of proton linked to microcirculation) were determined using prototype software (Siemens Healthineers). Pixel-based ADC maps were created based on monoexponential calculation from DWI with b-values of 0 and 800 s/mm2.
For DCE-MRI, unenhanced axial T1 weighted volumetric interpolated breath-hold examinations (VIBE) images were acquired for baseline T1 maps with flip angles of 2° and 15° before injecting contrast material. Images were then obtained immediately after a bolus injection of gadobutrol (Gadovist, Bayer Healthcare, Leverkusen, Germany) at a rate of 2 ml s−1 fora dose of 0.1 mmol/kg followed by a 20 ml normal saline flush. Total acquisition time was 5 min. The Tofts pharmacokinetic model was applied to calculate perfusion parameters including rate transfer constant between blood plasma and extracellular/extravascular space (Ktrans), volume fraction of extracellular/extravascular space (Ve), and rate intravasation constant (Kep) calculated asKtrans/Ve.17 Population-averaged arterial-input function (AIF) was applied with intermediate type.22 Semiquantitative parameters of the initial area under the time–signal intensity curve (iAUC) were also provided from the time-to-signal intensity curve as iAUC in 60 s.
Imaging analysis
MRI images were analyzed with region of interest (ROI) in a blind fashion by two musculoskeletal radiologists. Quantitative analysis of IVIM-DWI and DCE-MRI were obtained by manually placing a ROI within as much of the solid portion of the soft tissue tumor as possible, excluding cystic, hemorrhagic, or necrotic areas on the representative axial plane (Figure 2). Each ROI was placed on the T1 weighted images after gadolinium injection, and this ROI was propagated automatically on each of the parametric maps by software Syngo Via (Siemens Healthineers). The mean values of the parametric maps from the pixels within the ROIs were calculated. ROIs were visually corrected if any image distortions or artifacts were identified.
Figure 2.
ROI analysis. A, Axial fat-suppressed contrast-enhanced TSE T1 weighted image shows ROI placed manually within solid tumor portion on representative axial plane.B, IVIM-DWI parametric map and C, DCE-MRI parametric map show ROIs copied and pasted in the same place and shape on selective slice. DCE-MRI, dynamic contrast-enhanced magnetic resonance imaging; IVIM-DWI, intravoxel incoherent motion diffusion-weighted imaging; ROI, region of interest; TSE, turbo spine echo
Statistical analysis
χ2 test and Student’s t-test were used to compare the findings between the two groups. The Mann–Whitney U test was used to compare the findings for subgroup analysis.Receiver operating characteristic (ROC) analysis with area under the curve (AUC) was used to calculate sensitivity, specificity, and accuracy according to a defined cut-off value. Logistic regression analysis was performed to determine independent parameters for predicting a malignant lesion. Multiple ROC comparison was performed for prediction models representing the ROC AUC. Interobserver agreement for the measurement was evaluated by intraclass correlation coefficient (ICC) using a two-way mixed model with absolute agreement. Results were evaluated within a 95% confidence interval (CI), and significance level was set at p < 0.05. Statistical analysis was conducted using software packages (SPSS v. 20.0, Chicago, IL; and MedCalc, v. 12.7, Mariakerke, Belgium).
Results
Final diagnoses were established on the basis of histopathology, and the cases were composed of 35 benign and 32 malignant lesions. Detailed demographic and histopathological information of the two groups are described in Table 1. Interobserver agreement of IVIM-DWI was considered excellent for ADC and D (ICC = 0.95–0.98), good for D* and f(ICC = 0.58–0.65). For DCE-MRI, interobserver agreements for Ktrans, Kep and iAUC were considered good (ICC = 0.72–0.74) and that forVe was poor (ICC = 0.38).
Table 1.
Demographic and histologic information for benign and malignant soft tissue tumors
| Demographic data | Benign group (n = 35) | Malignant group (n = 32) | p value |
|---|---|---|---|
| Age (years) | 53 ± 16 | 58 ± 14 | 0.227 |
| Gender (m : f) | 16 : 19 | 14 : 18 | 0.872 |
| Histologic results (n) | Schwannoma (8) | Undifferentiated sarcoma (8) | |
| Hemangioma (6) | Synovial sarcoma (4) | ||
| Angioleiomyoma (4) | Myxoidliposarcoma (3) | ||
| Fibromatosis (4) | Myxofibrosarcoma (3) | ||
| Nodular fasciitis (3) | Lymphoma (3) | ||
| Fibroma (2) | Epithelioidhemangioendothelioma (2) | ||
| Glomus tumor (2) | Leiomyosarcoma (2) | ||
| Tenosynovial giant cell tumor (2) | Angiosarcoma (1) | ||
| Myxoma (2) | Low grade fibromyxoid sarcoma (1) | ||
| Neurofibroma (1) | Myeloid sarcoma (1) | ||
| Neuroma (1) | Extraskeletalmesenchymalchondrosarcoma (1) | ||
| Malignant melanoma (1) | |||
| Malignant peripheral nerve sheath tumor (1) | |||
| Malignant solitary fibrous tumor (1) |
Benign vs malignant soft tissue tumors
Among the IVIM-DWI parameters, ADC (1170 ± 488 µm2/s) and D (1132 ± 500 µm2/s) in malignant tumors were significantly lower than those of benign tumors (1472 ± 349 µm2/s and 1415 ± 374 µm2/s; p = 0.007 and 0.015, respectively) (Figures 3 and 4). There were no significant differences in D* and f between the two groups (Table 2). ROC analysis demonstrated that ADC and D showed the highest sensitivities of 74 and 71%, respectively (specificities of 73 and 83%, cut-offs of ≤1310 µm2/s and ≤1190 µm2/s indicating malignancy) with AUC values of 0.752 and 0.742 with statistical significance (p = 0.005 and 0.006) (Table 2).
Figure 3.
A 50-year-old male with myeloid sarcoma. A, Axial fat-suppressed contrast-enhanced TSE T1 weighted image shows 2.5 cm enhancing mass (arrows) along radial nerve in the elbow, mimicking schwannoma. B, Mass demonstrates high signal with hypointense portion (arrows) on DWI with b value of 800 s/mm2. C, Impeded water diffusivity (arrows) on ADC map. IVIM-DWI parameters and DCE-MRI parameters suggest malignant soft tissue tumor. This case was pathologically confirmed as myeloid sarcoma. D. Results of quantitative multiparametric MRI. ADC, apparent diffusion coefficient; DCE-MRI, dynamic contrast enhanced magnetic resonance imaging; DWI, diffusion-weighted imaging; IVIM-DWI, intravoxel incoherent motion diffusion-weighted imaging; TSE, turbo spin echo.
Figure 4.
A 71-year-old female with myxoma. A, Axial fat-suppressed contrast-enhanced TSE T1 weighted image shows 4 cm heterogeneously enhancing mass (arrows) in proximal thigh. B, Bright signal is not demonstrated on DWI with b value of 800 s/mm2. C, The mass is hyperintense (arrows) on ADC map. IVIM-DWI parameters and DCE-MRI parameters suggest a benign soft tissue tumor. The mass was pathologically confirmed as myxoma. D. Results of quantitative multiparametric MRI. ADC,apparent diffusion coefficient; DCE-MRI, dynamic contrast enhanced magnetic resonance imaging; DWI, diffusion-weighted imaging; IVIM-DWI, intravoxel incoherent motion diffusion-weighted imaging; TSE, turbo spin echo.
Table 2.
Comparison of parameters between benign and malignant soft tissue tumors
| Parameters | Benign group | Malignant group | p value | Cut-off values | Sensitivity (%) | Specificity (%) | Accuracy (%) | AUC |
|---|---|---|---|---|---|---|---|---|
| IVIM DWI | n = 30 | n = 31 | ||||||
| D (µm2/s) | 1415 ± 374 | 1132 ± 500 | 0.015a | ≤1190 | 71 (22/31) | 83 (25/30) | 77 (47/61) | 0.742a |
| Da (µm2/s) | 297 ± 96 | 258 ± 83 | 0.094 | ≤274 | 58 (18/31) | 57 (17/30) | 57 (35/61) | 0.638 |
| f (%) | 101 ± 58 | 83 ± 36 | 0.149 | ≤82 | 55 (17/31) | 57 (17/30) | 56 (34/61) | 0.576 |
| ADC (µm2/s) | 1472 ± 349 | 1170 ± 488 | 0.007a | ≤1310 | 74 (23/31) | 73 (22/30) | 74 (45/61) | 0.752a |
| DCE-MRI | n = 35 | n = 32 | ||||||
| Ktrans (min−1 × 103) | 92 ± 67 | 209 ± 160 | <0.001a | >110 | 81 (26/32) | 77 (27/35) | 79 (53/67) | 0.792a |
| Kep(min−1 × 103) | 311 ± 230 | 737 ± 488 | <0.001a | >368 | 78 (25/32) | 71 (25/35) | 75 (50/67) | 0.817a |
| Ve (%) | 44 ± 28 | 32 ± 17 | 0.043a | ≤35 | 69 (22/32) | 54 (19/35) | 61 (41/67) | 0.616 |
| iAUC (%) | 12 ± 9 | 23 ± 14 | <0.001a | >14 | 72 (23/32) | 69 (24/35) | 70 (47/67) | 0.771a |
AUC, area under the curve; DCE-MRI, Dynamic contrast enhanced magnetic resonance imaging; IVIM DWI, intravoxel incoherent motion diffusion-weighted imaging.
All data are means (±standard deviation).
indicates statistical significance.
Among the DCE-MRI parameters, Ktrans (209 ± 160 min−1x103), Kep (737 ± 488 min−1x103), and iAUC (23±14%) in malignant tumors were significantly higher than those in benign tumors (92 ± 67 min−1 x 103, 311 ± 230 min−1 x 103, and 12±9% respectively; p < 0.001 for all), while Ve (32±17%) in malignant tumors was significantly lower than that in benign tumors (44±28%, p = 0.043) (Table 2). ROC analysis found that Ktrans had the highest sensitivity of 81% (specificity of 77%, cut-off of >110 min−1x103 indicating malignancy) with an AUC of 0.792 with statistical significance (p < 0.001). Kepshowed the highest AUC of 0.817 with statistical significance (p < 0.001), with a value greater than 368 min−1x103 indicating malignancy (sensitivity of 78% and specificity of 71%). The iAUC also showed an AUC of 0.771 with statistical significance (p < 0.001), with a value greater than 14% indicating malignancy (sensitivity of 72% and specificity of 69%) (Table 2).
Subgroup analysis for myxoid soft tissue tumors
There were no significant differences in IVIM-DWI to differentiate between benign and malignant myxoid soft tissue tumors. There was no parameter in IVIM-DWI showing an AUC with statistical significance (Table 3).
Table 3.
Comparison and diagnostic performance of parameters between benign and malignant myxoid soft tissue tumors
| Parameter | Benign group | Malignant group | p- value | Cut-off values | Sensitivity (%) | Specificity (%) | Accuracy (%) | AUC |
|---|---|---|---|---|---|---|---|---|
| IVIM DWI | n = 11 | n = 8 | ||||||
| D (µm2/s) | 1623 ± 466 | 1687 ± 565 | 0.791 | ≤1416 | 50 (4/8) | 63 (7/11) | 57 (11/19) | 0.511 |
| Da (µm2/s) | 268 ± 38 | 232 ± 44 | 0.077 | ≤233 | 50 (4/8) | 81 (9/11) | 68 (13/19) | 0.739 |
| f (%) | 77 ± 46 | 68 ± 34 | 0.662 | ≤65.5 | 75 (6/8) | 54 (6/11) | 63 (12/19) | 0.545 |
| ADC (µm2/s) | 1656 ± 438 | 1720 ± 524 | 0.776 | ≤1407 | 50 (4/8) | 81 (9/11) | 68 (13/19) | 0.545 |
| DCE-MRI | n = 11 | n = 8 | ||||||
| Ktrans (min−1 × 103) | 63 ± 37 | 122 ± 73 | 0.035a | >110 | 62 (5/8) | 100 (11/11) | 84 (16/19) | 0.744 |
| Kep(min−1 × 103) | 262 ± 279 | 454 ± 279 | 0.158 | >220 | 87 (7/8) | 63 (7/11) | 73 (14/19) | 0.750a |
| Ve (%) | 57 ± 35 | 34 ± 14 | 0.100 | ≤43 | 87 (7/8) | 63 (7/11) | 73 (14/19) | 0.693 |
| iAUC (%) | 8 ± 5 | 17 ± 12 | 0.037a | >14 | 62 (5/8) | 90 (10/11) | 78 (15/18) | 0.761a |
AUC, area under the curve;DCE-MRI, Dynamic contrast enhanced magnetic resonance imaging; IVIM DWI, intravoxel incoherent motion diffusion-weighted imaging
All data are means (±standard deviation).
indicates the statistical significance.
Of the DCE-MRI parameters, Ktrans (122 ± 73 min−1x103) and iAUC (17±12%) in malignant myxoid soft tissue tumors were significantly higher than those in benign myxoid soft tissue tumors (63 ± 37 min−1x103 and 8±5% respectively; p = 0.035 and 0.037). ROC analysis showed Kepto have the highest sensitivity of 87% (specificity of 63%, cut-off of >220 min−1 × 103 indicating malignancy) with an AUC of 0.750 with statistical significance (p = 0.001). Better specificity was achieved with iAUC (specificity of 90%, sensitivity of 62%, cut-off of >14% indicating malignancy) with an AUC of 0.761 with statistical significance (p = 0.001) (Table 3).
Subgroup analysis for non-myxoid soft tissue tumors
The ADC (979 ± 300 µm2/s) and D (938 ± 296 µm2/s) in malignant non-myxoid soft tissue tumors were significantly lower than in benign non-myxoid soft tissue tumors (1365 ± 238 µm2/s and 1294 ± 249 µm2/s, respectively; p < 0.001). ROC analysis found that ADC and D showed the highest sensitivity of 87% (specificities of 73 and 79%, cut-offs of ≤1261 µm2/s and ≤1183 µm2/s indicating malignancy, respectively) with AUCs of 0.835 and 0.831 with statistical significance (p = 0.001) (Table 4).
Table 4.
Comparison and diagnostic performance of parameters between benign and malignant non-myxoid soft tissue tumors
| Parameter | Benign group | Malignant group | p value | Cut-off values | Sensitivity (%) | Specificity (%) | Accuracy (%) | AUC |
|---|---|---|---|---|---|---|---|---|
| IVIM DWI | n = 19 | n = 23 | ||||||
| D (µm2/s) | 1294 ± 249 | 938 ± 296 | <0.001a | ≤1183 | 87 (20/23) | 79 (15/19) | 83 (35/42) | 0.831a |
| Da (µm2/s) | 312 ± 114 | 266 ± 92 | 0.152 | ≤279 | 65 (15/23) | 57 (11/19) | 62 (26/42) | 0.634 |
| f (%) | 114 ± 61 | 88 ± 36 | 0.081 | ≤88 | 60 (14/23) | 68 (13/19) | 64 (27/42) | 0.638 |
| ADC (µm2/s) | 1365 ± 238 | 979 ± 300 | <0.001a | ≤1261 | 87 (20/23) | 73 (14/19) | 81 (34/42) | 0.835a |
| DCE-MRI | n = 24 | n = 24 | ||||||
| Ktrans (min−1 × 103) | 105 ± 72 | 238 ± 172 | 0.001a | >110 | 87 (21/24) | 66 (16/24) | 77 (37/48) | 0.798a |
| Kep(min−1 × 103) | 333 ± 207 | 832 ± 510 | <0.001a | >358 | 87 (21/24) | 66 (16/24) | 77 (37/48) | 0.855a |
| Ve (%) | 38 ± 22 | 31 ± 18 | 0.303 | ≤33 | 62 (15/24) | 50 (12/24) | 56 (24/48) | 0.584 |
| iAUC (%) | 14 ± 10 | 26 ± 15 | 0.002a | >12 | 87 (21/24) | 62 (15/24) | 75 (36/48) | 0.766a |
AUC, area under the curve; DCE-MRI, Dynamic contrast enhanced magnetic resonance imaging; IVIM DWI, intravoxel incoherent motion diffusion-weighted imaging.
All data are means (±standard deviation).
indicates the statistical significance.
Ktrans (238 ± 172 min−1x103), Kep (832 ± 510 min−1x103), and iAUC (26±15%) in malignant non-myxoid soft tissue tumors were significantly higher than in benign non-myxoid soft tissue tumors (105 ± 72 min−1x103, 333 ± 207 min−1x103, and 14±10% respectively; p = 0.001,<0.001, and 0.002). ROC analysis determined that Ktrans, Kep, and iAUC showed the same sensitivity of 87% (specificities of 66%, 66%, and 62%, cut-offs of >110 min−1x103, >358 min−1x103, and >12% indicating malignancy, respectively) with AUCs of 0.798, 0.855, and 0.766 with statistical significance (p < 0.001 for all)(Table 4).
Multivariate logistic regression analysis for predicting malignant soft tissue tumors
IVIM-DWI parameters were analyzed by stepwise multivariate logistic regression analysis, and D (OR, 0.998; 95% CI, 0.997–0.999) was an independent factor for predicting malignancy. The DCE-MRI with multivariate logistic regression analysis showed that Kep (OR, 1.004; 95% CI, 1.001–1.006) and iAUC (OR, 1.064; 95% CI, 1.001–1.131) were independent factors for predicting malignancy.
Three prediction models were designed as follows – first model, D alone; second model, D combined with Kep; third model, D combined with Kep and iAUC. ROC analyses of the three logistic regression models showed that the ROC AUC of predicted probability increased by adding parameters to D (AUC of first model, 0.750; second model, 0.809; third model, 0.838) without statistically significant differences (p = 0.415, 0.406, and 0.184 between first vs second, second vs third, and first vs third, respectively)(Figure 5).
Figure 5.
Graph showing ROC comparison between prediction models (D alone, D combined with Kep, and D combined with Kep and iAUC; AUC, 0.750, 0.809, 0.838, respectively) for predicting malignancy. (AUC,) area under the curve; ROC), receiver operating characteristic curve
False negative and false positive cases
D demonstrated a significantly high AUC value among IVIM-DWI parameters, and the D cut-off value resulted in seven false positive cases (two angioleiomyoma, two nodular fasciitis, fibromatosis, a fibroma, and a tenosynovial giant cell tumor and eight false negative cases (three myxoid liposarcomas,a fibromyxoidsarcoma, a malignant peripheral nerve sheath tumor, an epithelioid hemangioendothelioma, a leiomyosarcoma, anda synovial sarcoma).
Among DCE-MRI parameters, Ktrans and Kep revealed significantly high AUC values. Based on Ktrans cut-off value, there were eight false positive cases (two hemangiomas, two glomus tumors, a fibromatosis, a nodular fasciitis, a tenosynovial giant cell tumor, and an angioleiomyoma) and six false negative cases (two myxoid liposarcomas, a synovial sarcoma, an epithelioid hemangioendothelioma, a leiomyosarcoma, and a low-grade fibromyxoid sarcoma), while based on Kep cut-off value, there were 10 false positive cases (three hemangiomas, two nodular fasciitis cases, a schwannoma, a glomus tumor, a myxoma, a neuroma, and a tenosynovial giant cell tumor) and 7 false negative cases (two myxoid liposarcomas, a myxofibrosarcoma, a malignant peripheral nerve sheath tumor, a malignant solitary fibrous tumor, an epithelioid hemangioendothelioma, and a synovial sarcoma).
Discussion
Our study showed that D, ADC, Ktrans, Kep, and iAUC can be used to differentiate benign and malignant soft tissue tumors. Only the DCE-MRI parameters of Ktrans, and iAUC enabled differentiation of benign and malignant myxoid soft tissue tumors, whereas D, ADC, Ktrans, Kep, and iAUC can be used to differentiate benign and malignant non-myxoid soft tissue tumors.
Although we did not assess capacity of conventional MRI to characterize soft tissue tumors, Chung et al23 reported the diagnostic values of conventional MRI using depth, size, and heterogeneity to differentiate between benign and malignant soft tissue tumors, resulting in a sensitivity of 64%, a specificity of 85%, and an accuracy of 77%. By comparing the results of that study using conventional MRI by Chung et al23 and the results presented in this study using multiparametric MRI, we can infer that addition of multiparametric MRI to conventional MRI can improve the diagnostic performance in distinguishing malignant from benign soft tissue tumors, as it increased the sensitivity (71 to 81%) while maintaining similar levels of specificity (69 to 83%) and accuracy (70 to 79%).
D and ADC were significantly different between malignant and benign soft tissue tumors with good diagnostic performance, consistent with previous studies.12,13 We expected that D would be better than ADC for differentiating soft tissue tumors because D eliminates the contributions of tissue perfusion to reflect tissue diffusivity more precisely than ADC.24 However, the AUCs of ADC and D were similar based on ROC analysis (ADC, 0.752; D, 0.742) and multivariate logistic regression analysis, in agreement with the recent study by Lim et al.13 Our results were contrary to the previous study by Rijswijk et al.6 Rijswijk et al6 investigated soft tissue tumors using early IVIM-DWI with five b-values (0–701 s/mm2) at 1.5 T and reported that D was significantly different between benign and malignant soft tissue tumors, whereas ADC was not. Early IVIM-DWI with fewer b-values at 1.5 T might be one of the reasons for the difference. We also assume that microcapillary perfusion might be heterogeneous or higher in benign soft tissue tumors than in malignant soft tissue tumors, which seems to contribute to there being no significant difference between ADC and D for differentiating soft tissue tumors. This assumption is reinforced by our result that there were no differences in D* and f between malignant and benign soft tissues, similar to the previous study by Lim et al..13
We found significant differences in Ktrans, Kep, and iAUC between malignant and benign soft tissue tumors with good diagnostic performances. Our results are similarly to that of Leplat et al,19 who found only Ktrans was only different between malignant and benign soft tissue tumors. However, we observed some discrepancies in Kep, iAUC, and Ktrans, which might be due to different AIF selection and ROI measurement techniques. Leplat el al19 used automatic software AIF selection, and ROIs were placed on the tumor areas that were most intensely enhancing on the AUC map. Our study found that among quantitative DCE-MRI parameters, Kephad the highest AUC by ROC analysis (Kep, 0.817; Ktrans, 0.792; iAUC, 0.771) and multivariate logistic regression analysis, similar to a previous study.25 Oto et al25 found the best correlation between quantitative perfusion parameters and histologic angiogenesis parameters were noted between Kep and microvessel density. Kep is an index of Ktrans / Ve, and the compounding effect of Ktrans and Ve may contribute to better correlation of Kep and microvessel density.25
Our study found that soft tissue tumors are distinguished better when evaluated with a combination of IVIM-DWI and DCE-MRI than when evaluated by IVIM-DWI alone. The subgroup analysis of our study classified benign and malignant soft tissue tumors into two groups each: myxoid and non-myxoid subgroups, and showed that ADC and D were not significantly different between benign and malignant myxoid soft tissue tumors. Although a classification of myxoid soft tissue tumors based on the WHO classification26 did not include peripheral nerve sheath tumors in myxoid soft tissue tumors, we categorized peripheral nerve sheath tumors as myxoid soft tissue tumors because they histologically contained a small myxoid component.26,27 As illustrated by the subgroup analysis results, IVIM-DWI has limitations in differentiation of soft tissue tumors because ADC is affected by the extracellular matrix as well as by cellularity.9,21,27 Regardless of benign or malignant behavior, soft tissue tumors contain varying degrees of myxoid matrix, which causes overlap in ADC values between benign and malignant soft tissue tumors. This was demonstrated by subgroup analysis in our study and previous studies.9,21,27 The ADC and D of malignant myxoid soft tissue tumors were slightly higher than those of benign myxoid soft tissue tumors, although there was no significant difference (p > 0.05). We thought these results were affected by inclusion of peripheral nerve sheath tumors in the subgroup of myxoid soft tissue tumors. As schwannoma, which has a relatively small myxoid component, was included as a benign subgroup, the ADC and D of the benign subgroup were demonstrated to be lower than those of the malignant subgroup.
Our study had several limitations. First, it was conducted retrospectively at a single institution with a limited study population. Although we enrolled consecutive patients with clear inclusion criteria, the possibility of selection bias remains. Second, the study population and range of tumors were relatively small because surgeons at our institution (Seoul St. Mary's Hospital) do not perform surgery on tumors that are definitely benign on imaging studies. Third, ROIs may contain calcifications of soft tissue tumors that affect ADC values because calcifications cannot be completely excluded on MRI. Finally, we did not assess conventional MRI because it is subjective and our purpose was to provide an objective quantitative value to differentiate benign and malignant soft tissue tumors in clinical practice. Since the usefulness of IVIM-DWI and DCE-MRI in addition to conventional MRI was not been addressed in our study, it is not known how useful is multiparametric MRI for diagnosis of soft tissue tumors compared to conventional MRI alone. We expect this evaluation to be performed in future studies.
In summary, quantitative analysis of multiparametric MRI including IVIM-DWI and DCE-MRI enables better differentiation of benign and malignant soft tissue tumors. Particularly in myxoid tumors, the DCE-MRI parameters of Ktrans, and iAUC enable differentiation of benign and malignant myxoid soft tissue tumors.
Footnotes
Conflict of interest: The authors of this manuscript declare no relationships with any companies, whose products or services may be related to the subject matter of the article.
Funding: The authors state that this work has not received any funding.
Contributor Information
Seul Ki Lee, Email: beneffy@catholic.ac.kr.
Won-Hee Jee, Email: whjee12@gmail.com.
Chan Kwon Jung, Email: ckjung@catholic.ac.kr.
Yang-Guk Chung, Email: ygchung@catholic.ac.kr.
REFERENCES
- 1.Kransdorf MJ, Jelinek JS, Moser RP, Utz JA, Brower AC, Hudson TM, et al. Soft-Tissue masses: diagnosis using MR imaging. AJR Am J Roentgenol 1989; 153: 541–7https://doi.org/10.2214/ajr.153.3.541. doi: 10.2214/ajr.153.3.541 [DOI] [PubMed] [Google Scholar]
- 2.Crim JR, Seeger LL, Yao L, Chandnani V, Eckardt JJ. Diagnosis of soft-tissue masses with MR imaging: can benign masses be differentiated from malignant ones? Radiology 1992; 185: 581–6. doi: 10.1148/radiology.185.2.1410377 [DOI] [PubMed] [Google Scholar]
- 3.Benedikt RA, Jelinek JS, Kransdorf MJ, Moser RP, Berrey BH. Mr imaging of soft-tissue masses: role of gadopentetate dimeglumine. J Magn Reson Imaging 1994; 4: 485–90. doi: 10.1002/jmri.1880040341 [DOI] [PubMed] [Google Scholar]
- 4.Moulton JS, Blebea JS, Dunco DM, Braley SE, Bisset GS, Emery KH. Mr imaging of soft-tissue masses: diagnostic efficacy and value of distinguishing between benign and malignant lesions. AJR Am J Roentgenol 1995; 164: 1191–9. doi: 10.2214/ajr.164.5.7717231 [DOI] [PubMed] [Google Scholar]
- 5.Gielen JLMA, De Schepper AM, Vanhoenacker F, Parizel PM, Wang XL, Sciot R, et al. Accuracy of MRI in characterization of soft tissue tumors and tumor-like lesions. A prospective study in 548 patients. Eur Radiol 2004; 14: 2320–30. doi: 10.1007/s00330-004-2431-0 [DOI] [PubMed] [Google Scholar]
- 6.van Rijswijk CSP, Kunz P, Hogendoorn PCW, Taminiau AHM, Doornbos J, Bloem JL. Diffusion-Weighted MRI in the characterization of soft-tissue tumors. J Magn Reson Imaging 2002; 15: 302–7. doi: 10.1002/jmri.10061 [DOI] [PubMed] [Google Scholar]
- 7.Einarsdóttir H, Karlsson M, Wejde J, Bauer HCF. Diffusion-Weighted MRI of soft tissue tumours. Eur Radiol 2004; 14: 959–63. doi: 10.1007/s00330-004-2237-0 [DOI] [PubMed] [Google Scholar]
- 8.Jeon JY, Chung HW, Lee MH, Lee SH, Shin MJ. Usefulness of diffusion-weighted MR imaging for differentiating between benign and malignant superficial soft tissue tumours and tumour-like lesions. Br J Radiol 2016; 89: 20150929https://doi.org/10.1259/bjr.20150929. doi: 10.1259/bjr.20150929 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 9.Lee S-Y, Jee W-H, Jung J-Y, Park MY, Kim S-K, Jung C-K, et al. Differentiation of malignant from benign soft tissue tumours: use of additive qualitative and quantitative diffusion-weighted MR imaging to standard MR imaging at 3.0 T. Eur Radiol 2016; 26: 743–54. doi: 10.1007/s00330-015-3878-x [DOI] [PubMed] [Google Scholar]
- 10.Song Y, Yoon YC, Chong Y, Seo SW, Choi Y-L, Sohn I, et al. Diagnostic performance of conventional MRI parameters and apparent diffusion coefficient values in differentiating between benign and malignant soft-tissue tumours. Clin Radiol 2017; 72: 691.e1–691.e10https://doi.org/10.1016/j.crad.2017.02.003. doi: 10.1016/j.crad.2017.02.003 [DOI] [PubMed] [Google Scholar]
- 11.Le Bihan D, Breton E, Lallemand D, Aubin ML, Vignaud J, Laval-Jeantet M. Separation of diffusion and perfusion in intravoxel incoherent motion MR imaging. Radiology 1988; 168: 497–505. doi: 10.1148/radiology.168.2.3393671 [DOI] [PubMed] [Google Scholar]
- 12.Wu H, Zhang S, Liang C, Liu H, Liu Y, Mei Y, et al. Intravoxel incoherent motion MRI for the differentiation of benign, intermediate, and malignant solid soft-tissue tumors. J Magn Reson Imaging 2017; 46: 1611–8. doi: 10.1002/jmri.25733 [DOI] [PubMed] [Google Scholar]
- 13.Lim HK, Jee W-H, Jung J-Y, Paek MY, Kim I, Jung C-K, et al. Intravoxel incoherent motion diffusion-weighted MR imaging for differentiation of benign and malignant musculoskeletal tumours at 3 T. Br J Radiol 2018; 91: 20170636https://doi.org/10.1259/bjr.20170636. doi: 10.1259/bjr.20170636 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 14.van Rijswijk CSP, Geirnaerdt MJA, Hogendoorn PCW, Taminiau AHM, van Coevorden F, Zwinderman AH, et al. Soft-Tissue tumors: value of static and dynamic gadopentetate dimeglumine-enhanced MR imaging in prediction of malignancy. Radiology 2004; 233: 493–502. doi: 10.1148/radiol.2332031110 [DOI] [PubMed] [Google Scholar]
- 15.Park MY, Jee W-H, Kim SK, Lee S-Y, Jung J-Y. Preliminary experience using dynamic MRI at 3.0 Tesla for evaluation of soft tissue tumors. Korean J Radiol 2013; 14: 102–9. doi: 10.3348/kjr.2013.14.1.102 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 16.Noebauer-Huhmann I-M, Amann G, Krssak M, Panotopoulos J, Szomolanyi P, Weber M, et al. Use of diagnostic dynamic contrast-enhanced (DCE)-MRI for targeting of soft tissue tumour biopsies at 3T: preliminary results. Eur Radiol 2015; 25: 2041–8. doi: 10.1007/s00330-014-3576-0 [DOI] [PubMed] [Google Scholar]
- 17.Tofts PS, Brix G, Buckley DL, Evelhoch JL, Henderson E, Knopp MV, et al. Estimating kinetic parameters from dynamic contrast-enhanced T(1)-weighted MRI of a diffusable tracer: standardized quantities and symbols. J Magn Reson Imaging 1999; 10: 223–32. doi: [DOI] [PubMed] [Google Scholar]
- 18.Marzi S, Stefanetti L, Sperati F, Anelli V. Relationship between diffusion parameters derived from intravoxel incoherent motion MRI and perfusion measured by dynamic contrast-enhanced MRI of soft tissue tumors. NMR Biomed 2016; 29: 6–14. doi: 10.1002/nbm.3446 [DOI] [PubMed] [Google Scholar]
- 19.Leplat C, Hossu G, Chen B, De Verbizier J, Beaumont M, Blum A, et al. Contrast-Enhanced 3-T perfusion MRI with quantitative analysis for the characterization of musculoskeletal tumors: is it worth the trouble? AJR Am J Roentgenol 2018; 211: 1092–8. doi: 10.2214/AJR.18.19618 [DOI] [PubMed] [Google Scholar]
- 20.Hong JH, Jee W-H, Jung C-K, Jung J-Y, Shin SH, Chung Y-G. Soft tissue sarcoma: adding diffusion-weighted imaging improves MR imaging evaluation of tumor margin infiltration. Eur Radiol 2019; 29: 2589–97. doi: 10.1007/s00330-018-5817-0 [DOI] [PubMed] [Google Scholar]
- 21.Maeda M, Matsumine A, Kato H, Kusuzaki K, Maier SE, Uchida A, et al. Soft-Tissue tumors evaluated by line-scan diffusion-weighted imaging: influence of myxoid matrix on the apparent diffusion coefficient. J Magn Reson Imaging 2007; 25: 1199–204. doi: 10.1002/jmri.20931 [DOI] [PubMed] [Google Scholar]
- 22.Orton MR, d'Arcy JA, Walker-Samuel S, Hawkes DJ, Atkinson D, Collins DJ, et al. Computationally efficient vascular input function models for quantitative kinetic modelling using DCE-MRI. Phys Med Biol 2008; 53: 1225–39. doi: 10.1088/0031-9155/53/5/005 [DOI] [PubMed] [Google Scholar]
- 23.Chung WJ, Chung HW, Shin MJ, Lee SH, Lee MH, Lee JS, et al. Mri to differentiate benign from malignant soft-tissue tumours of the extremities: a simplified systematic imaging approach using depth, size and heterogeneity of signal intensity. Br J Radiol 2012; 85: e831–6. doi: 10.1259/bjr/27487871 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 24.Sigmund EE, Cho GY, Kim S, Finn M, Moccaldi M, Jensen JH, et al. Intravoxel incoherent motion imaging of tumor microenvironment in locally advanced breast cancer. Magn Reson Med 2011; 65: 1437–47. doi: 10.1002/mrm.22740 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 25.Oto A, Yang C, Kayhan A, Tretiakova M, Antic T, Schmid-Tannwald C, et al. Diffusion-Weighted and dynamic contrast-enhanced MRI of prostate cancer: correlation of quantitative Mr parameters with Gleason score and tumor angiogenesis. AJR Am J Roentgenol 2011; 197: 1382–90. doi: 10.2214/AJR.11.6861 [DOI] [PubMed] [Google Scholar]
- 26.Baheti AD, Tirumani SH, Rosenthal MH, Howard SA, Shinagare AB, Ramaiya NH, et al. Myxoid soft-tissue neoplasms: comprehensive update of the taxonomy and MRI features. AJR Am J Roentgenol 2015; 204: 374–85. doi: 10.2214/AJR.14.12888 [DOI] [PubMed] [Google Scholar]
- 27.Nagata S, Nishimura H, Uchida M, Sakoda J, Tonan T, Hiraoka K, et al. Diffusion-Weighted imaging of soft tissue tumors: usefulness of the apparent diffusion coefficient for differential diagnosis. Radiat Med 2008; 26: 287–95. doi: 10.1007/s11604-008-0229-8 [DOI] [PubMed] [Google Scholar]