Features of MR Imaging that Differentiate between Immunohistochemically Diagnosed Dedifferentiated Liposarcoma and Myxoid Liposarcoma

Abstract

Purpose

This study aimed to compare the differences in the imaging findings for dedifferentiated liposarcoma (DDLS) and myxoid liposarcoma (MLS).

Methods

The study included 30 patients with histopathologically confirmed DDLS and 13 patients with MLS. All DDLSs and MLSs were diagnosed immunohistochemically using MDM2 and DDIT3 staining, respectively. Conventional MRI, CT, and ¹⁸F-fluorodeoxyglucose-positron emission tomography/CT findings were retrospectively evaluated and compared between the 2 pathologies.

Results

The median age of patients with DDLS was higher than that of patients with MLS (74 vs. 46 years, P < 0.01). In 10 DDLSs and 7 MLSs with fatty areas, the well-differentiated liposarcoma-like fatty areas were more common in DDLS than in MLS (70% vs. 14%), whereas septal/linear fatty areas were less common in DDLS than in MLS (30% vs. 86%, P < 0.05). The T2-hyperintense area of non-fatty area was less common in DDLS than in MLS (50% vs. 92%, P < 0.05), and the tumor-to-muscle signal intensity ratio of non-fatty areas on T2-weighted images was lower in DDLS than in MLS (3.18 vs. 5.92, P < 0.01). Apparent diffusion coefficient value was lower in DDLS than in MLS (1.29 vs. 2.10 × 10⁻³mm²/sec, P < 0.01). Unenhanced CT attenuation of non-fatty area was greater in DDLS than in MLS (33 vs. 19 Hounsfield unit, P < 0.01).

Conclusion

MRI features are valuable in differentiating MLS from DDLS. Younger age, septal/linear fatty areas, and high signal intensity of non-fatty areas on T2-weighted images were useful for diagnosing MLS.

Introduction

Liposarcoma is classified into 5 different histological subtypes: atypical lipomatous tumor/well-differentiated liposarcoma (ALT/WDLS), dedifferentiated liposarcoma (DDLS), myxoid liposarcoma (MLS), pleomorphic liposarcoma (PLS), and myxoid pleomorphic liposarcoma (MPLS).^1–3 According to the most recent study, including the largest number of 12822 liposarcoma cases, the frequencies of WDLS, DDLS, and MLS are 41%, 28%, and 22%, respectively.⁴ PLS and MPLS are rare variant, and the frequencies of PLS and MPLS are reported to be 5%–10% and <1%, respectively.

DDLS is caused by the transition from an ALT/WDLS to a sarcoma of variable histological grade. DDLS and ALT/WDLS are characterized at the molecular level by the amplification of genes within 12q13–15, including MDM2 and CDK4.^2,5 MLS comprises uniform, round-to-ovoid cells and varying numbers of small lipoblasts in a prominent myxoid stroma with a chicken-wire vascular network. Translocations that produce FUS–DDIT3 or, less commonly, EWSR1–DDIT3 fusion transcripts are pathognomonic.³

MLS is relatively sensitive to radiotherapy and chemotherapy than other soft tissue sarcomas; thus, surgery and chemoradiotherapy are the standard treatments for MLS.⁶ The standard treatment for localized DDLS includes surgery and/or radiotherapy, whereas that for advanced DDLS includes an anthracycline-based regimen.⁶ In addition, as MLS has a characteristic tendency to metastasize to extrapulmonary sites, particularly to the spine and soft tissues,^6,7 whole body MRI can identify extrapulmonary metastases that are occult on CT or fluorodeoxyglucose (FDG)-positron emission tomography (PET)/CT⁸ (Fig. 1). The local recurrence, distant metastasis, and 5-year overall survival rates in DDLS and MLS were 16.7%–47.6% and 15.1%–24%, 10.2%–37.5% and 5.8%–31%, and 30%–40% and 67%–90%, respectively.^9–12 Because of the significant differences in local recurrence, distant metastasis, and 5-year overall survival rates between the 2 pathologies, we believe that the accurate differentiation between DDLS and MLS enables estimate prognosis. DDLS and MLS have different treatment strategies, metastatic sites, and prognosis; therefore, an accurate differential diagnosis between DDLS from MLS is required to provide appropriate treatment.

Fig. 1.

A 33-year-old woman with MLS with bone and soft tissue metastasis. Sagittal T1-weighted image (a) showing hypointense vertebral metastases in T12 and L4 (arrows) and axial T1-weighted image in L4 (b) showing hypointense vertebral metastasis with extension into the spinal canal (arrows). Sagittal CT (c) and axial CT in L4 (d) showing nearly normal spine, indicating intertrabecular bone metastasis. (e) Axial CT image showing subcutaneous metastasis on the left side of the chest. MLS, myxoid liposarcoma.

On MRI, DDLS typically shows a mass composed of fatty and non-fatty components, often without fatty components.^13,14 MLS commonly appears high signal intensity on T2-weighted images due to abundant myxoid stroma and exhibits liner or amorphous foci of fat.¹⁵ DDLS can resemble MLS due to the marked hyperintensity on T2-weighted images caused by abundant myxoid stroma or extensive necrosis and may have liner fat foci as with MLS. Meanwhile, lipid-rich MLS can resemble DDLS or WDLS.¹⁶ Although ALT/WDLS can be distinguished from other subtypes of liposarcoma by the abundant amount of fatty component,¹⁷ MRI findings between DDLS and MLS considerably overlap. PLS and MPLS are very rare subtypes of liposarcoma, and they could not be included in this study due to the small number of populations.

Many studies have described the imaging findings of MLS^16,18–21 and difference of imaging findings between MLS and myxoma,^22,23 and some have reviewed the imaging findings of immunohistochemically diagnosed DDLS by MDM2.^13,14,17 By contrast, no studies have evaluated the imaging findings of immunohistochemically diagnosed MLS using DDIT3; therefore, there is concern that fat-poor MLS mimicking other myxoid sarcoma might not be included as research subjects in the previous imaging studies regarding MLS. In addition, there has been no comparative study of imaging findings of DDLS and MLS. Therefore, this study aimed to improve the ability of differential diagnosis through comparing the imaging findings of histopathologically and immunohistochemically diagnosed DDLS and MLS.

Materials and Methods

Patients

The study was approved by the Human Research Committee of our hospital’s institutional review board, and the requirement for written informed consent was waived off by the board. The study was conducted in accordance with the Health Insurance Portability and Accountability Act of 1996. Patients with histopathologically and immunohistochemically confirmed DDLS and MLS who underwent preoperative imaging were retrospectively studied from April 2014 to October 2023. Eight patients with DDLS in the retroperitoneal region were excluded because MLS rarely occurs in here²⁴ and all patients with retroperitoneal DDLS did not undergo MRI. This study included 30 patients with DDLS (18 men and 12 women; age range, 47–89 years; median age, 74 years) and 13 patients with MLS (9 men and 4 women; age range, 25–79 years; median age, 46 years) in the musculoskeletal region. The locations of DDLS and MLS were thigh in 16 and 9, lower leg in 2 and 2, upper arm in 3 and 1, and trunk in 9 and 1, respectively. DDLS and MLS were diagnosed according to the 5th edition of the World Health Organization classification of soft tissue tumors in 2020. In the present study, all cases of DDLS and MLS were positive for MDM2 and DDIT3 (>10% nuclear immunoreactivity), respectively.²⁵

MRI parameters

MRI images were performed with 1.5T (Intera Achieva 1.5T Pulsar or Inginea prodiva 1.5T CS; Philips Healthcare, Best, the Netherlands) in 31 patients (22 with DDLSs and 9 with MLSs) or 3.0T units (Intera Achieva 3.0T Quasar Dual or Inginea 3.0T CX; Philips Healthcare) in the remaining 12 patients (8 with DDLSs and 4 with MLSs). All MRI images were performed under the following conditions: section thickness of 3–5 mm, intersection gap of 1–2 mm, and FOV of 24 × 24 to 43 × 43 cm. Axial nonfat-suppressed T2-weighted (TR/TE, 3000–6153/80–95 ms), axial and coronal or sagittal fat-suppressed T2-weighted (TR/TE, 2000–6026/60–80 ms), and axial and coronal or sagittal nonfat-suppressed T1-weighted (TR/TE, 382–761/7.6–12 ms) were obtained in all patients. In 27 patients (20 with DDLS and 7 with MLS), axial diffusion-weighted imaging with a single shot spin-echo echo-planar (TR/TE, 3000–6400/51–82 ms; b-value = 0 and 1000 s/mm²) were obtained. Diffusion-weighted images were obtained with 1.5T in 19 patients (14 with DDLSs and 5 with MLSs) or 3.0T units in the remaining 8 patients (6 with DDLSs and 2 with MLSs). In 36 patients (25 with DDLS and 11 with MLS), contrast-enhanced images were obtained with an intravenous injection of 0.1 mmol/kg of gadopentetate dimeglumine (Magnevist; Bayer HealthCare, Leverkusen, Germany) or gadobutrol (Gadavist; Bayer HealthCare) using T1-weighted fat-suppressed sequences (TR/TE, 479–652/11.8–15.5 ms).

CT imaging

We used a 16- (LightSpeed 16; GE Healthcare, Milwaukee, WI, USA) or 64-slice CT scanners (Brilliance CT 64; Philips Healthcare, or Discovery CT 750HD, GE Healthcare). Unenhanced CT images of 37 patients (27 with DDLS and 10 with MLS) were obtained. Coronal or sagittal reconstructed images of 2.5 mm were obtained.

¹⁸F-fluorodeoxyglucose-PET/CT protocols

Eighteen patients (14 with DDLS and 4 with MLS) underwent whole-body PET/CT (Biograph Sensation 16; Siemens Medical Solutions, Malvern, PA, USA or Discovery MI.x; GE Healthcare, Milwaukee, WI, USA). After at least 5 hours of fasting, patients underwent an intravenous injection of ¹⁸F-FDG. CT and whole-body PET were performed approximately 60 minutes after the administration of ¹⁸F-FDG. The technical parameters of the 16-multidetector CT were as follows: gantry rotation speed 0.5 s; table speed per gantry rotation 24 mm; and quiet-breathing data acquisition. Transverse images were reconstructed with 2 mm section thickness and no overlap.

Image analysis

Two radiologists (with 25 and 11 years of experience in the interpretation of musculoskeletal imaging), who were unaware of clinical information and pathological diagnosis, independently reviewed all images. Any disagreements were resolved through consensus. The evaluation items were cited from the previous imaging studies regarding DDLS and MLS.^17,21,23

The reviewers first assessed the qualitative MRI findings in terms of depth (superficial or deep), configuration (round/oval or lobulated), margin (well-defined or ill-defined), tail sign (presence or absence), flow void (presence or absence), and presence or absence of fatty and non-fatty areas. Superficial location refers to the subcutaneous tissue and skin above the superficial investing fascia, which separates the subcutaneous tissue layer from the underlying muscle. The margins between the tumor and surrounding normal tissue were evaluated. The tail sign was defined as a linear or tapered extension along the dermis on fat-suppressed T2-weighted images. Flow void was defined as punctate or tortuous signal loss within the tumor on T2-weighted images. Fatty areas were defined as a hyperintense area similar to subcutaneous fat on T1- and T2-weighted images and with a lower signal intensity on fat-suppressed T2-weighted images. In addition, we confirmed CT images in terms of the presence of fat density (<0 Hounsfield unit [HU]) in patients who underwent CT. If fatty areas were present, they were classified as WDLS-like (predominant fat areas with thickened septa) or septal/linear (predominant non-fatty areas with marbled texture fat).²¹ If both WDLS-like and septal/linear fatty areas were present, the reviewer determined the predominant type. On T1-weighted imaging, non-fatty areas were defined as an isointense area (>1 cm) relative to the skeletal muscles.¹⁷

Second, if non-fatty areas were present, T2-hyointense and T2-hyperintense areas (presence or absence), predominant signal of non-fatty areas on T2-weighted images (low, intermediate, or high), homogeneity on T1- and T2-weighted images (homogeneous or heterogeneous), unenhanced area of non-fatty area (presence or absence), degree of enhancement of non-fatty area (mild or marked), and enhancement pattern of non-fatty area (diffuse, peripheral, or nodular) were evaluated. T2-hyointense and T2-hyperintense areas were defined as isointense areas relative to skeletal muscles and subcutaneous fat or water, respectively, on T2-weighted images.¹⁷ Predominant signal intensities of non-fatty areas on nonfat-suppressed T2-weighted images were classified into low, intermediate, and high, which were isointense compared to the skeletal muscle, lymph node, and subcutaneous fat or water, respectively. Diffuse, peripheral, and nodular enhancement patterns were defined as extensive enhancement (>90% area of the tumor), peripheral dominant enhancement (the circumference ≥ 180°), and central dominant or focal peripheral enhancement (the circumference < 180°), respectively. The presence or absence of calcification and fat attenuation on CT was evaluated.

Finally, the radiologists determined the maximum diameter of the whole tumor, signal intensity ratio of non-fatty areas on T1-, T2-, and contrast-enhanced T1-weighted images, apparent diffusion coefficient (ADC) value of non-fatty areas, unenhanced CT attenuation of non-fatty areas, and maximum standardized uptake value (SUVmax) of the whole tumor. The signal intensity of non-fatty areas on T1-, T2-, and contrast-enhanced T1-weighted images was measured by defining ROIs. ROIs on T1-, T2-, and contrast-enhanced T1-weighted images were placed as widely as possible within non-fatty areas on a single cross-sectional image at the level of the maximum tumor diameter. The reviewer also assessed the signal intensity of the skeletal muscles at the same level as the lesion and calculated the tumor-to-muscle signal intensity ratio. ADC values of non-fatty areas were measured on ADC maps by overlying an ROI over the lesion. ROIs on ADC maps were placed as widely as possible within non-fatty areas using the respective T2-weighted images as references. Quantitative values were calculated by the average of the radiologists.

Statistical analysis

All statistical analyses were conducted using EZR (Saitama Medical Center, Jichi Medical University, Saitama, Japan), a graphical user interface for R (The R Foundation for Statistical Computing, Vienna, Austria).²⁶ The Mann–Whitney U test and Fisher’s exact test were used to compare quantitative and qualitative data between DDLS and MLS, respectively. Receiver operating characteristics (ROC) curve analysis was used to determine the diagnostic performance and the optional cutoff value for differentiating MLS from DDLS. Odds ratio and 95% confidence intervals (CIs) of predicting factors for diagnosing MLS were calculated by logistic regression analysis. Subsequently, the multiple regression model equation was constructed. A P value less than 0.05 was considered to be significant. Kappa statistics and the intraclass correlation coefficient were used to determine interobserver agreement in qualitative and quantitative assessments, respectively.

Results

Table 1 presents the clinical and qualitative imaging findings for DDLS and MLS. There was no difference in terms of gender, location, depth, configuration, margin, tail sign, and flow void between DDLS and MLS. Fatty areas were noted in 10 and 7 patients with DDLS and MLS, respectively (30% vs. 54%, P = 0.31). DDLS had more WDLS-like fatty areas than MLS (70% vs. 14%), whereas septal/linear fatty areas were less common in DDLS than in MLS (30% vs. 86%, P < 0.05) (Figs. 2, 3, 4, 5). Three cases of DDLS with septal/linear fatty areas did not have WDLS-like fatty areas, but all of them had histologically confirmed well-differentiated components. Meanwhile, 1 case of MLS with WDLS-like fatty areas also had septal/linear fatty areas. Non-fatty areas were present in both DDLS and MLS; however, T2-hyperintense areas were less common in DDLS than in MLS (50% vs. 92%, P < 0.05). The predominant signal of non-fatty areas in DDLS and MLS on T2-weighted images was low in 10% and 0%, intermediate in 57% and 15% and high in 33% and 85% of the patients, respectively (P < 0.01) (Figs. 2, 3, 4, 5). There was no difference in T2-hypointense areas and heterogeneity on T1- and T2-weighted images between DDLS and MLS. On contrast-enhanced T1-weighted images, the degree of enhancement of non-fatty area in DDLS and MLS was mild at 32% and 77% and marked at 68% and 23%, respectively (P < 0.05). The enhancement patterns of non-fatty area in DDLS and MLS were diffuse in 16% and 18%, peripheral in 80% and 27%, nodular in 4% and 55%, respectively (P < 0.05). On CT images, there was no difference in calcification and fat attenuation between DDLS and MLS. Fatty areas were observed in 17 cases (10 DDLS and 7 MLS) on MRI. Among them, fatty areas on CT were observed in all 13 cases (8 DDLSs and 5 MLS) who underwent CT imaging.

Table 1.

Clinical and qualitative imaging findings of DDLS and MLS

	DDLS (n = 30)	MLS (n = 13)	P value	Kappa value
Gender – Male	18 (60)	9 (69)	0.74
Location – Thigh	16 (53)	9 (69)	0.33
Depth – Superficial	10 (33)	5 (38)	0.74	0.79
Configuration – Lobulated	20 (67)	8 (62)	0.74	0.68
Margin – Well-defined	22 (73)	13 (100)	0.082	0.40
Tail sign	24 (80)	9 (69)	0.46	0.11
Flow void	11 (37)	5 (38)	>0.99	0.49
Fatty area	10 (30)	7 (54)	0.31	0.76
Well-differentiated liposarcoma-like	7 (70)	1 (14)	0.025*	1.00
Septal/linear	3 (30)	6 (86)
Non-fatty area	30 (100)	13 (100)	>0.99	0.80
T2-hypointense area	15 (50)	4 (31)	0.32	0.40
T2-hyperintense area	15 (50)	12 (92)	0.014*	0.74
Predominant signal of non-fatty area on T2WI			0.009*	0.69
Low signal intensity	3 (10)	0 (0)
Intermediate signal intensity	17 (57)	2 (15)
High signal intensity	10 (33)	11 (85)
Homogeneous on T1WI	7 (23)	3 (23)	>0.99	0.72
Homogeneous on T2WI	2 (7)	2 (15)	0.74	0.63
CET1WI	n = 25	n = 11
Non-enhanced area on non-fatty area	20 (80)	11 (100)	0.30	0.31
Degree of enhancement of non-fatty area			0.034*	0.48
Mild	8 (32)	8 (73)
Marked	17 (68)	3 (27)
Pattern of enhancement of non-fatty area			<0.001	0.50
Diffuse	4 (16)	2 (18)
Peripheral	20 (80)	3 (27)
Nodular	1 (4)	6 (55)
CT	n = 27	n = 10
Calcification	4 (15)	3 (30)	0.36	0.80
Fat attenuation	8 (30)	5 (50)	0.28	0.71

Qualitative data are numbers of patients with percentages in parentheses.

^*Significant difference was observed between DDLS and MLS (P < 0.05).CET1WI, contrast-enhanced T1-weighted images; DDLS, dedifferentiated liposarcoma; MLS, myxoid liposarcoma; T1WI, T1-weighted images; T2WI, T2-weighted images.

Fig. 2.

A 78-year-old man with DDLS in the thigh. (a) Axial T1-weighted image showing an intramuscular mass with non-fatty (arrow) and well-differentiated liposarcoma-like fatty areas (asterisks). (b) Axial fat-suppressed contrast-enhanced T1-weighted image showing mild enhancement (arrow). (c) Axial T2-weighted image showing an intramuscular mass with non-fatty (arrow) and fatty areas (asterisks). Non-fatty areas exhibiting heterogeneously intermediate signal intensity (ROI; tumor-to-muscle signal intensity ratio = 2.71). Axial diffusion-weighted image (d) and ADC map (e) showing restricted diffusion (arrow) with low ADC value (0.82 × 10^-3mm²/sec). (f) Axial unenhanced CT showing an intramuscular mass with non-fatty (arrow) and fatty areas (asterisks). Non-fatty areas showing soft tissue attenuation (42 HU). (g) A resected specimen of DDLS (hematoxylin and eosin staining; scale bar, 250 µm) showing high cellularity tumor with pleomorphism and fibrous stroma. ADC, apparent diffusion coefficient; DDLS, dedifferentiated liposarcoma; HU, Hounsfield unit.

Fig. 3.

A 64-year-old woman with DDLS in the thigh. (a) Axial T1-weighted image showing intramuscular mass (arrow) comprising only non-fatty areas. (b). Axial fat-suppressed contrast-enhanced T1-weighted image showing homogeneously marked enhancement (arrow). (c) Axial T2-weighted image showing an intramuscular mass (arrow) comprising only non-fatty areas. Non-fatty areas exhibiting heterogeneously intermediate signal intensity (ROI; Tumor-to-muscle signal intensity ratio = 2.5). Axial diffusion-weighted image (d) and ADC map (e) showing restricted diffusion (arrow) with relatively ADC value (1.02 × 10⁻³mm²/sec). (f) ¹⁸F-FDG PET/CT image showing high FDG uptake (SUVmax = 22.9). ADC, apparent diffusion coefficient; DDLS, dedifferentiated liposarcoma; FDG, fluorodeoxyglucose; PET, positron emission tomography; SUVmax, maximum standardized uptake value.

Fig. 4.

A 37-year-old man with MLS in the thigh. (a) Axial T1-weighted image showing an intramuscular mass (arrow) mainly comprising non-fatty area with septal/linear fatty areas (arrowheads). (b) Axial fat-suppressed contrast-enhanced T1-weighted image showing minimal enhancement (arrow) with fat-suppressed areas (arrowheads). (c) Axial T2-weighted image showing an intramuscular mass (arrow) with septations. Non-fatty areas showing markedly high signal intensity (ROI; tumor-to-muscle signal intensity ratio = 5.45). (d) Axial diffusion-weighted image showing high signal intensity (arrow). (e) ADC map showing high ADC value (arrow, 2.09 × 10⁻³mm²/sec). (f) Axial unenhanced CT showing an intramuscular mass (arrow) with partially fatty areas (arrowheads). Non-fatty areas showing low attenuation (22 HU). (g) ¹⁸F-FDG PET/CT image showing low FDG uptake (SUVmax = 3.5). (h) A resected specimen of MLS (hematoxylin and eosin staining; scale bar, 250 µm) showing low cellularity tumor with mature fat and myxoid stroma. ADC, apparent diffusion coefficient; FDG, fluorodeoxyglucose; HU, Hounsfield unit; MLS, myxoid liposarcoma; PET, positron emission tomography; SUVmax, maximum standardized uptake value.

Fig. 5.

A 25-year-old woman with MLS in the thigh. (a) Axial T1-weighted image showing an intramuscular mass (arrow) comprising non-fatty area alone. (b) Axial fat-suppressed contrast-enhanced T1-weighted image showing mild enhancement (arrow). (c) Axial T2-weighted image showing an intramuscular mass (arrow) with homogeneously and markedly high signal intensity (ROI; Tumor-to-muscle signal intensity ratio = 8.48). (d) Axial unenhanced CT showing an intramuscular mass (arrow) comprising only non-fatty areas. Non-fatty areas showing low attenuation (12 HU). HU, Hounsfield unit; MLS, myxoid liposarcoma.

Table 2 presents the clinical and quantitative imaging findings of DDLS and MLS. The median age was higher in DDLS than in MLS (74 vs. 46 years, P < 0.01). On T2-weighted images, the signal intensity ratio of non-fatty areas relative to skeletal muscles was lower in DDLS than in MLS (3.18 vs. 5.92, P < 0.01) (Figs. 2, 3, 4, 5). There was a significant difference in ADC values (1.29 vs. 2.10 × 10⁻³mm²/sec, P < 0.01), unenhanced CT attenuation of non-fatty areas (33 vs. 19 HU, P < 0.01), and SUVmax of the whole tumor (7.36 vs. 3.50, P < 0.01) between DDLS and MLS. There was no difference in the maximum diameter and signal intensity ratio of non-fatty areas on T1- and contrast-enhanced T1-weighted images.

Table 2.

Clinical and quantitative imaging findings of DDLS and MLS

	DDLS (n = 30)	MLS (n = 13)	P value	ICC
Age (year)	74.0 [65.0–80.0]	46.0 [37.0–65.0]	<0.001*	NA
Maximum diameter (mm)	69.5 [51.0–120]	62.0 [51.0–150]	0.87	0.98
Non-fatty area
SIR on T1WI	1.09 [1.01–1.25]	1.20 [1.03–1.35]	0.25	0.75
SIR on T2WI	3.18 [2.59–4.95]	5.92 [5.15–7.02]	<0.001*	0.91
SIR on CET1WI	2.07 [1.76–2.43] (n = 25)	1.70 [1.43–1.93] (n = 11)	0.086	0.35
ADC value (×10−3mm2/sec)	1.29 [1.00–1.64] (n = 20)	2.10 [2.03–2.33] (n = 7)	0.005*	0.97
Unenhanced CT attenuation (HU)	32.5 [25.8–38.5] (n = 27)	19.3 [12.8–21.5] (n = 10)	0.003*	0.84
SUVmax of whole tumor	7.36 [5.55–12.2] (n = 14)	3.50 [3.26–3.82] (n = 4)	0.006*	NA

Quantitative data are expressed as medians with interquartile in square brackets.

^*Significant difference was observed between DDLS and MLS (P < 0.05).ADC, apparent diffusion coefficient; CET1WI;  contrast-enhanced T1-weighted images; DDLS, dedifferentiated liposarcoma; HU,  Hounsfield unit; MLS, myxoid liposarcoma; NA, not applicable; SIR, signal intensity ratio; SUVmax , maximum standardized uptake value; T1WI, T1-weighted images; T2WI, T2-weighted images.

Table 3 summarizes the results of ROC curve analysis for differentiating MLS from DDLS. The AUC values for age, signal intensity ratio of non-fatty areas relative to skeletal muscles on T2-weighed images, ADC value, unenhanced CT attenuation, and SUVmax of the whole tumor were 0.83, 0.82, 0.85, 0.82, and 0.97, respectively.

Table 3.

Results of ROC curve analysis for differentiating MLS from DDLS

	AUC (95% CI)	Cutoff value	Sensitivity	Specificity	P value
Age	0.83 (0.69–0.98)	≤48 years	61.5	93.3	<0.001*
SIR of non-fatty area on T2WI	0.82 (0.70–0.95)	≥3.56	100	63.3	<0.001*
ADC value (n = 27)	0.85 (0.67–1.00)	≥1.97 × 10−3mm2/sec	85.7	90.0	0.007*
Unenhanced CT attenuation (n = 37)	0.82 (0.66–0.98)	≤22 HU	80.0	85.2	0.003*
SUVmax of whole tumor (n = 18)	0.97 (0.91–1.00)	≤4.77	100	85.7	<0.001*

^*Significant differences were observed (P < 0.05).ADC,  apparent diffusion coefficient; AUC,  area under the curve; CI,  confidence interval; DDLS,  dedifferentiated liposarcoma; MLS,  myxoid liposarcoma; ROC, receiver operating characteristics; SIR, signal intensity ratio; SUVmax, maximum standardized uptake value; T2WI,  T2-weighted images.

Table 4 summarizes the results of univariate and multivariate analyses of clinical and imaging findings for predicting MLS. On univariate analysis, significant factors to predict MLS were age, septal/linear fatty area, predominant high signal of non-fatty area on T2-weighted images, signal intensity ratio of non-fatty area relative to skeletal muscle on T2-weighted image, ADC value, and unenhanced CT attenuation. Among similar MRI findings, predominant high signal of non-fatty area on T2-weighted images was excluded from multivariate analysis because predominant high signal of non-fatty area on T2-weighted images had greater P value than signal intensity ratio of non-fatty area relative to skeletal muscle on T2-weighted images. ADC value and unenhanced CT attenuation were also excluded from multivariate analysis because DWI value and unenhanced CT were not performed in all patients. Consequently, multivariate analysis revealed that age and septal/linear fatty area were independent predictors for MLS. Odds ratio of age and septal/linear fatty area were 0.90 (95% CI 0.84–0.97) and 16.9 (95% CI 1.55–184), respectively.

Table 4.

Univariate and multivariate analyses of clinical and imaging findings for predicting MLS

	Univariate analysis		Multivariate analysis
	Odds ratio (95% CI)	P value	Odds ratio (95% CI)	P value
Age	0.91 (0.86–0.96)	0.001*	0.90 (0.84–0.97)	0.009*
Septal/linear fatty area
Absent	Reference	Reference
Present	7.71 (1.53–38.8)	0.013*	16.9 (1.55–184)	0.020*
Predominant signal of non-fatty area on T2WI
Low or intermediate signal intensity	Reference	Reference	NA
High signal intensity	11.0 (2.04–59.4)	0.005*	NA
SIR of non-fatty area on T2WI	1.80 (1.21–2.68)	0.004*	1.32 (0.81–2.16)	0.27
ADC value (n = 27)	18.5 (1.78–193)	0.015*	NA
Unenhanced CT attenuation (n = 37)	0.87 (0.78–0.96)	0.006*	NA
SUVmax of whole tumor (n = 18)	NA		NA

^*Significant differences were observed (P < 0.05).ADC,  apparent diffusion coefficient; CI,  confidence interval; MLS, myxoid liposarcoma; NA, not available; SIR, signal intensity ratio; SUVmax, maximum standardized uptake value; T2WI , T2-weighted images.

Table 5 summarizes the results of univariate and multivariate analyses of clinical and imaging findings for predicting MLS in patients (n = 24) who underwent diffusion-weighted imaging and unenhanced CT. On univariate analysis, significant factors to predict MLS were age, signal intensity ratio of non-fatty area relative to skeletal muscle on T2-weighted images, ADC value, and unenhanced CT attenuation. On multivariate analysis, there was no significant factor for predicting MLS.

Table 5.

Subanalysis of univariate and multivariate analyses of clinical and imaging findings for predicting MLS in patients who underwent diffusion-weighted imaging and unenhanced CT

	Univariate analysis		Multivariate analysis
	Odds ratio (95% CI)	P value	Odds ratio (95% CI)	P value
Age	0.90 (0.83–0.98)	0.013*	0.94 (0.85–1.04)	0.22
Septal/linear fatty area
Absent	Reference	Reference
Present	4.50 (0.48–42.2)	0.19
Predominant signal of non-fatty area on T2WI
Low or intermediate signal intensity	Reference	Reference
High signal intensity	7.50 (0.73–76.8)	0.09
SIR of non-fatty area on T2WI	2.46 (1.24–4.88)	0.0099*	1.29 (0.44–3.77)	0.64
ADC value	10.4 (1.22–88.5)	0.032*	1.65 (0.05–49.5)	0.77
Unenhanced CT attenuation	0.84 (0.73–0.97)	0.018*	0.94 (0.78–1.12)	0.46
SUVmax of whole tumor (n = 18)	NA

ADC,  apparent diffusion coefficient; CI,  confidence interval; MLS, myxoid liposarcoma; NA,  not available; SIR, signal intensity ratio; SUVmax, maximum standardized uptake value; T2WI,  T2-weighted images.

^*Significant differences were observed (P < 0.05).

Table 6 shows the results of multiple linear regression analysis for predicting MLS. This proposed model showed that age (coefficient = −0.012, P < 0.001) and septal/linear fatty area (coefficient = 0.364, P = 0.007) were significant factors for diagnosing MLS. The multiple linear regression model equation for predicting MLS is:

$$\begin{array}{cc}Y=& 0.788-0.012\times \text{Age + 0}\text{.364}\\ & \times \text{Septal / linear}\text{fatty}\text{area}\text{(present}\text{=}\text{1,}\text{absent}=0)\\ & +0.057\\ & \times \text{SIR}\text{of}\text{non}-\text{fatty}\text{area}\text{on}\text{T2}-\text{weighted}\text{images}\end{array}$$

Table 6.

Results of multiple linear regression analysis for predicting MLS

	Coefficient	P value
(Intercept)	0.788
Age	−0.012	<0.001
Septal/linear fatty area	0.364	0.007
SIR on T2WI	0.057	0.063

MLS, myxoid liposarcoma; SIR, signal intensity ratio; T2WI, T2-weighted images.

ROC curve analysis of the above formula using age, septal/linear fatty area, and SIR on T2WI revealed that the AUC value was 0.92 (95% CI 0.82–1.00).

Discussion

Our study revealed MRI features of immunohistochemically diagnosed DDLS and MLS. DDLS had more WDLS-like fatty areas than MLS, whereas septal/linear fatty areas were less frequently observed in DDLS than in MLS. Moreover, signal intensity ratio of non-fatty areas relative to skeletal muscles on T2-weighed images and ADC value of non-fatty areas were lower in DDLS than in MLS. Unenhanced CT attenuation and SUVmax of non-fatty area were higher in DDLS than in MLS. Multiple linear regression analysis revealed that age and septal/linear fatty area were significant factors for diagnosing MLS.

Age was younger in MLS than in DDLS in this study. According to previous studies, MLS has a peak incidence in the 4th to 5th decades,^3,15 whereas DDLS has in the 7th to 8th decades of life.^13,14 The development of the disease at a young age is a characteristic of MLS compared to DDLS.

Fatty components have been reported to be detectable on MRI in 63%–75% of MLS cases.^18,20,23 However, because these previous studies did not perform immunostaining, they might not include fat-poor MLS mimicking other myxoid sarcoma. Meanwhile, the present study enrolled immunohistochemically diagnosed MLS alone; therefore, the frequency of fatty components within MLS was relatively lower in this study than in previous studies.

In the present study, the frequencies of WDLS-like and septal/linear fatty areas were different between DDLS and MLS. Previous studies reported that fatty areas on MRI could be detected in 27%–61%^13,14,17 and 49%–93%^{16,18,20,21,23} cases of DDLS and MLS, respectively. These results are consistent with the results of our study. Because DDLS involves the progression from a WDLS, the fatty components of WDLS are visible on MRI.^13,14 On T1-weighted images, MLS primarily presents with low signal intensity owing to non-fatty areas with septal or marbled high signal intensity owing to textual pattern of fat.^18,21 Histopathologically, MLS is characterized by an abundance of myxoid stroma and varying amounts of small lipoblasts or mature fat.³ Therefore, the differences in the appearances of fatty areas can help differentiate DDLS from MLS.

In the present study, the T2-hyperintense area was more common in MLS, with the majority of MLS cases showing this area. Furthermore, on T2-weighted images, the predominant signal intensity of in non-fatty area of MLS was high, and the signal intensity ratio was higher in MLS than in DDLS. In a previous study that examined MRI findings of DDLS, the predominant signal intensity relative to skeletal muscles on T2-weighted images was high in 59% (13/22) and intermediate in 41% (9/22) of the patients.¹⁷ Previous MRI studies on MLS revealed that 75% (24/32) cases had a myxoid component of >60%,¹⁸ and 80% (24/30) of the cases had a myxoid component fraction of ≥50%.²⁰ The non-fatty component of DDLS includes various histological subtypes of sarcoma and most frequently resembles undifferentiated pleomorphic sarcoma or intermediate to high-grade myxofibrosarcoma, which can contain myxoid stroma or necrosis.^2,3 The signal intensity of non-fatty components is useful in differentiating between the 2 pathologies; however, radiologists should know that DDLS can have T2-hyperintense areas due to myxoid stroma or necrosis.

In this study, unenhanced CT attenuation of non-fatty area was higher in DDLS, but ADC value of non-fatty area was lower in DDLS. In previous studies, CT attenuations of DDLS and MLS were 26.7 and 13.8–14.0 HU, respectively.^23,27 ADC values for MLS ranged from 1.97 to 2.51 × 10⁻³mm²/sec.^19,22 Although no study has reported ADC values for DDLS, it is estimated to be low because of the increased cellularity of the high-grade sarcoma component. Because MLS contains abundant myxoid stroma, which results in lower cellularity, it has a low CT value but a high ADC value. DDLS had high CT attenuation and low ADC value, which helped differentiate it from MLS.

In this study, SUVmax of the whole tumor was higher in DDLS. In previous studies, the SUVmax of DDLS ranged from 9.3 to 16,^27,28 whereas that of MLS ranged from 3.0 to 5.21.^23,27,28 MLS typically has much lower metabolic activity than other malignancies and exhibits low FDG accumulation, which may show faint uptake close to background levels, and can overlap with benign intramuscular myxoma, unless MLS contains a significant number of round cells.²³ In contrast, DDLS, like other malignancies, is hypermetabolic.²⁷ Because FDG-PET/CT is clinically useful to search for distant metastasis, the use of FDG-PET/CT is usually recommended for clinical staging before treatment. If the primary tumor exhibits low FDG uptake, MLS is more likely than DDLS; therefore, additional whole body MRI should be performed to search for bone metastasis. Therefore, SUVmax of the lesion can be useful in distinguishing between DDLS and MLS and selecting proper metastatic survey.

Some discrepancies were observed in results between Tables 4 and 5. Specifically, significant difference in septal/linear fatty area was not observed in univariate analysis (Table 5). Only 6 patients with MLS underwent both diffusion-weighted imaging and unenhanced CT. Among them, septal/linear fatty area was observed in only 2 patients. We believe that this discrepancy was caused by the disproportionate number of populations.

The study has several limitations. First, we included a relatively small number of cases, particularly in MLS cases. Second, because patients with retroperitoneal DDLS were excluded from this study, the difference in the location could not be evaluated sufficiently; however, patients with retroperitoneum MLS were not identified in our institution. Third, the quantitative findings (signal intensity ratio, ADC value, and CT attenuation) were evaluated using a single slice image and would not reflect the characteristics of entire tumor. Fourth, contrast-enhanced CT features could not be evaluated because only 10 patients (8 DDLSs and 2 MLSs) underwent contrast-enhanced CT. The degree of enhancement of non-fatty areas would be weaker in MLS than in DDLS on contrast-enhanced CT due to the abundant myxoid matrix within MLS. Fifth, although ROC curve analysis revealed that SUVmax was the most significant factor for diagnosing MLS from DDLS, SUVmax could not be available for univariate and multivariate analysis due to the small number of patients who underwent FDG-PET/CT. Finally, diffusion-weighted images were obtained using MRI scanners with varying magnetic field strengths (1.5 or 3T). However, regardless of the different magnetic field strengths, the difference in ADC values between DDLS and MLS was indisputable.

In conclusion, MRI findings are valuable in differentiating MLS from DDLS. WDLS-like fatty areas were common in DDLS, whereas septal/linear fatty areas were common in MLS. Furthermore, DDLS had a lower ADC value and higher unenhanced CT attenuation. Meanwhile, younger age, higher signal intensity of non-fatty areas on T2-weighted images caused by myxoid stroma was characteristic of MLS. The accurate knowledge of imaging findings for differentiating DDLS from MLS would lead to proper preoperative diagnosis and metastatic survey, resulting in selecting appropriate treatment strategy.