Role of chemical shift imaging in characterizing musculoskeletal soft tissue tumors

INTRODUCTION

Magnetic resonance imaging (MRI) plays a vital role in the assessment and diagnosis of musculoskeletal soft tissue tumors. The role of MRI is not only limited to delineating the anatomical extent of the tumor but also provides details on the composition of the tumor. This often allows a skilled radiologist to grade the aggressiveness of the tumor, which is very important for pre and post-treatment evaluation. Some of the functional sequences that are used in the musculoskeletal system include chemical shift imaging (CSI), diffusion-weighted imaging, perfusion imaging, diffusion tensor imaging, and MR spectroscopy.^[1] CSI is a technique of fat suppression and is routinely used for the detection and characterization of abdominal lesions (hepatic, renal, and adrenal).^[2-4] In the musculoskeletal system, CSI is well established for the characterization of localized and diffuse bone marrow lesions, mainly in the spine.^[5,6] In CSI, in-phase and opposed-phased imaging sequences are acquired based on the principle of detecting different precessional frequencies of water and fat to detect areas of fatty tissue. When fat and water protons are present in the same voxel and they are imaged in in-phase sequence, their signals will add up and increase the overall signal intensity of that particular voxel, while in opposed-phase sequence, their signal will cancel each other out and will result in an overall decrease in signal intensity.

In the spine, CSI is used to differentiate fatty marrow replacement from pathological marrow infiltration. Therefore, in case of fatty marrow replacement, signal intensity in opposed-phase sequence will reduce, whereas in pathological marrow infiltration, this will not be the case.^[7] The role of CSI has been studied in the past in cases of adrenal and renal tumors.^[8-10] Some authors have also evaluated the role of CSI in assessing musculoskeletal tumors.^[11-13] However, data regarding the characterization of soft tissue tumors using CSI are limited.

CSI can be applied in characterizing musculoskeletal soft tissue lesions based on the same principle to detect the presence of fat cells and forms the hypothesis of our study. The primary objective of our study was to determine the extent of signal drop in histopathological proven cases of soft tissue tumors of the musculoskeletal system. Our secondary objective was to determine the diagnostic accuracy of CSI in predicting the histopathological diagnosis of musculoskeletal soft tissue tumor.

MATERIAL AND METHODS

This retrospective cross-sectional study was conducted at a tertiary orthopedic oncology center over 1-year period. Local ethical committee approval was obtained. Our institutional oncology database was retrospectively searched with keywords “lump,” “oncology,” and “soft tissue mass.” Inclusion criteria were patients with histopathological diagnosed musculoskeletal soft tissue tumors with availability of MRI having CSI done within 6 months of biopsy. Patients with an MRI performed more than 6 months from the date of biopsy and those without a CSI sequence were excluded. Infective, post-traumatic and postoperative cases were also excluded. All images were acquired on Siemens 3T Skyra or Siemens 1.5T Sola (Erlangen, Germany). The parameters for CSI were field of view 340, Repetition Time (TR) 137, TE in phase −2.38 ms, TE out of phase −4.87 ms, slice thickness 3 mm, and flip angle 70°.

Applying inclusion and exclusion criteria, a total of 80 musculoskeletal soft tissue tumors that were discussed in our regional sarcoma MDT over the past 9 months were identified and collected for analysis. Based on the histopathological diagnosis, patients were grouped in 5 cohorts (lipomatous, lipomatous variant, neural, myxoid, and other soft tissue tumors).

RESULTS

Based on the histopathological diagnosis, 80 patients were grouped in 5 cohorts (lipomatous, lipomatous variant, neural, myxoid, and other soft tissue tumors). There were 44 males and 36 females with an average age 51.2 (± 17.4) years. The lipoma cohort included 22 patients, the lipoma variant cohort included 8 patients, 11 neural tumors, the myxoid tumor cohort had 8 patients, and the other soft-tissue tumors cohort included 31 patients. The mean signal drop in the lipoma cohort was −12.4 % (±5.9), in the lipoma variant cohort was −27.4% (±18.5), in the neural tumors cohort was −2.7% (±5.2), in the myxoid tumors cohort was −1.9 % (±5.2), and in the other soft tissue tumor cohort was 0.8 % (±18.3). A summary of mean measurements is provided in [Table 1 and Figures 1-5].

Close

Figure 1:

Lipoma. (a) Sagittal T1, (b) short tau inversion recovery, (c) in phase, and (d) opposed phase MR images showing subcutaneous lipoma (white arrow in a,b) that demonstrates signal drop off of 46 (3.6%).

Export to PPT

Close

Figure 2:

Spindle cell lipoma. (a) Sagittal T1, (b) short tau inversion recovery, (c) in phase, and (d) opposed phase MR images showing subcutaneous spindle cell lipoma (white arrow in a,b) that demonstrates signal drop off of 107 (66%).

Export to PPT

Close

Figure 3:

Hibernoma. (a) Axial T1, (b) short tau inversion recovery, (c) in phase, and (d) opposed phase MR images showing hibernoma (black arrow in a, white arrow in b) that demonstrates signal drop off of 70 (18.7%).

Export to PPT

Close

Figure 4:

Myxoma. (a) Axial T1, (b) short tau inversion recovery, (c) in phase, and (d) opposed phase MR images showing intramuscular (white arrow in a, b) that does not demonstrate any signal drop off.

Export to PPT

Close

Figure 5:

Melanoma metastasis. (a) Axial T1, (b) short tau inversion recovery, (c) in phase, and (d) opposed phase MR images showing subcutaneous melanoma metastasis (white arrow in a,b) that demonstrates a signal increase of 13 (6.9%).

Export to PPT

The cohort did not differ significantly in terms of age (P = 0.801). The cohort differed significantly in terms of in-phase signal intensity, out-of-phase signal intensity, and signal drop percentage on CSI (P < 0.001). Post hoc analysis demonstrated that lipoma variant dropout was statistically significantly higher than that observed in the neural tumor (P = 0.002), myxoid (P = 0.004), and other cohorts (P < 0.001). There was also a statistically significant difference between the lipoma and other cohorts (P = 0.008). Other pairs did not demonstrate a statistically significant difference; pair values are provided in Table 2.

Signal dropout demonstrated good ability to identify lipoma variants with an area under curve (AUC) of 0.85 (95% confidence interval [CI] = 0.69–0.99) [Figure 6]. A dropout of greater than 14% had a 75% sensitivity and 77.78% specificity.

Close

Figure 6:

Area under curve (AUC) for the ability of the percentage dropout to discriminate between lipoma variant cases and other soft-tissue tumors. AUC = 0.8438.

Export to PPT

A small number of lesion types consistently demonstrated increased signal on out-of-phase imaging compared to in-phase imaging [Figure 5]. The lesions with this finding were 4 cases of tenosynovial giant cell tumor (20 ± 14.4%), 3 cases of pilomatricomas (19 ± 8.5%), and one case of melanoma metastasis (6.9%).

ICC 2, 1 was used to calculate inter- and intraobserver reliability. Overall, there was excellent intra-observer reliability (ICC = 0.996, 95% C.I. 0.990–0.998) and inter-observer reliability (ICC 0.990, 95% C.I. 0.971–0.996).

DISCUSSION

CSI is one of the functional sequences used in musculoskeletal imaging for the differentiation of pathological marrow infiltration from normal marrow, red marrow hyperplasia, and reactive marrow edema. The role of CSI in characterizing musculoskeletal soft tissue tumors has not been assessed previously. The primary objective of our study was to determine the extent of signal drop-out in histopathological proven cases of soft tissue tumors of the musculoskeletal system. Our secondary objective was to determine the diagnostic accuracy of CSI in predicting histopathological diagnosis of musculoskeletal soft tissue tumor.

Based on the histopathological diagnosis, 80 patients in our study were grouped in 5 cohorts (lipomatous, lipomatous variant, neural, myxoid, and other soft tissue tumors) with a male-to-female ratio of 1.2 and average age 51.2 (±17.4) years. Lipoma is relatively easily to diagnose on conventional sequences involving T1 and fat suppression sequences without the aid of CSI. Lipoma variants can mimic soft tissue sarcoma and, in most scenarios, require either image-guided biopsy or excision biopsy for diagnosis. CSI can aid in the diagnosis of lipoma variants. The mean signal drop was maximum in lipoma variant cohort, followed by lipoma cohort, neural tumors cohort, myxoid tumors cohort, and other soft tissue tumor cohort. The cohort differed significantly in terms of in-phase signal intensity, out-of-phase signal intensity, and signal drop-out percentage on CSI. The signal drop off was higher in lipoma variant in comparison to lipoma which could be due to fat water interfaces which are seen in lipoma variants and not in lipomas. Signal dropout demonstrated good ability to identify lipoma variants from other musculoskeletal soft tissue tumors. Notably, some tumors such as tenosynovial giant cell tumor, melanoma metastasis, and pilomatricomas demonstrated increased signal on outof-phase imaging. The aforementioned findings are novel and have not been previously identified in published work.

Shannon et al., evaluated the added value of CSI, dynamic contrast-enhanced imaging, and diffusion-weighted imaging to conventional MRI for characterizing indeterminate lipomatous tumors. They retrospectively evaluated 32 patients with histopathologically proven lipomatous lesions and categorized the tumors on MRI with the above sequences in 3 groups (benign, intermediate/atypical lipomatous tumor, and malignant/dedifferentiated liposarcoma). They found that the diagnostic accuracy of the MRI did not improve with the addition of CSI, dynamic contrast-enhanced imaging, and diffusion-weighted imaging to conventional sequences. The authors found some potentially useful imaging features to differentiate different histopathological grades on conventional sequences and these features were the presence of thick septations and nodules.^[14] We did not evaluate the role of CSI in differentiating benign from malignant tumor group and conventional MRI features such as thick septa, nodular components, enhancement, and margins are more reliable features that can help in this regard.

Nishida et al., aimed to differentiate between different types of lipomas on imaging, but the study did not utilize advanced MRI sequences.^[15] ICC 2,1 was used to calculate inter- and intraobserver reliability and there was excellent intra-observer reliability and inter-observer reliability in our study.

Limitations of this work include its retrospective nature and small numbers of specific soft-tissue tumors. Prospective studies with larger sample sizes, focusing on lipoma variants and tenosynovial tumors should be performed to validate the role of CSI. Macroscopic fat, hemorrhagic/proteinaceous content, and artefacts can confound CSI interpretation. CSI performance varies by field strength (1.5 T vs. 3 T), and future studies can be planned to see the difference between the two commonly used field strengths in affecting CSI values and their role in differentiating tumor types.

CONCLUSION

CSI demonstrates promise in the evaluation of soft tissue lumps in particular lipoma variants. This is the first study to describe the features of soft-tissue musculoskeletal tumor on CSI sequences.