Magnetic resonance imaging assessing the correlation of components and prognosis in myxoid liposarcoma

Abstract

Background

Myxoid liposarcoma (MLS) is a subtype of liposarcoma characterized by its myxoid stroma and adipocyte differentiation. MLS is prone to recurrence and metastasis. Magnetic resonance imaging (MRI) plays a crucial role in evaluating tumor characteristics, enabling accurate diagnosis, and predicting patient prognosis.

Purpose

To analyze the components of MLS by MRI features and assess their correlation with prognosis.

Material and Methods

A total of 20 patients with MLS who underwent MRI were retrospectively included. Tumor components were analyzed by MRI features, and their prognostic correlation was assessed. Patients were divided into good and poor prognosis groups based on postoperative follow-up.

Results

The proportions of non-fatty/non-myxoid components in the good and poor prognosis groups were 15.00% (range = 10.00%–20.00%) and 70.00% (range = 52.50%–77.50%), respectively (P < 0.001). The proportion of myxoid composition also differed significantly between the two groups (75.00%, [range = 65.00%–85.00%] vs. 25.00% [range = 17.50%–42.50%]; P < 0.001). The good prognosis group had a greater mean apparent diffusion coefficient (ADC) value (1.66 ± 0.23 × 10⁻³ mm²/s) and a lower mean ADC low signal ratio (5.00% [range = 0%–10.00%]) in the non-fatty/non-myxoid areas than the poor group (1.21 ± 0.41 × 10⁻³ mm²/s; 20.00% [range = 11.00%–39.00%]; P = 0.006 and P = 0.001). The differences in the percentages of patients with a component ratio <25% and >50% in both the non-fatty/non-myxoid and myxoid groups were significant (P < 0.001 and P = 0.005).

Conclusion

Imaging features were closely associated with the histological components of MLS. The use of MRI features for assessing MLS components has important implications for prognostic prediction.

Introduction

Myxoid liposarcoma (MLS) is a rare malignant myxoid soft-tissue sarcoma composed of a capillary network, a mucus matrix, and spindle adipoblasts. MLS is the second most common subtype of liposarcomas, accounting for approximately 30% of all types of liposarcomas (1–5). This subtype has a 10-year survival rate of approximately 77%, with 14%–32% of cases metastasizing to locations such as deep muscles of the limbs (6,7). According to the 2020 World Health Organization (WHO) classification of soft-tissue tumors, MLS is divided into high- (>5%) and low-grade (<5%) based on the proportion of round cell components (8). The presence of a higher round cell component ratio in MLS significantly influences the risk of recurrence, metastasis, and mortality, thus increasing the likelihood of these adverse outcomes (6,7,9–12). Zhao et al. (13 considered peritumoral enhancement as a credible magnetic resonance imaging (MRI) feature for high-grade soft-tissue sarcoma, while MLS exhibited frequent well-demarcated boundaries with rare peritumoral infiltration, and capsule formation is commonly observed even in high-grade cases. Lowenthal et al. (14–16) found that the MRI characteristics of MLS were associated with the proportion of tumor components. In addition, the proportions of different components were correlated with tumor grade.

In this study, we assessed the role of MRI features in analyzing the components of MLS and evaluating their correlation with prognosis to provide an objective basis for clinical diagnosis and treatment decisions.

Material and Methods

Patients

In total, 106 patients were pathologically diagnosed with liposarcomas at our center (Jinhua Hospital, Zhejiang University School of Medicine) between January 2012 and March 2023. The main inclusion criteria for this study were as follows: (i) patients with histologically confirmed MLS; and (ii) patients whose MRI scans were acquired before surgery. The exclusion criteria were as follows: (i) MRI scans were unable to resolve or quantify tumor components; and (ii) pathological diagnostic MLS mixed with other sarcomas. Following the above criteria, a retrospective cohort was subsequently formed, consisting of 20 patients (12 men, 8 women; mean age = 54.6 years). The patients presented at the hospital with an asymptomatic physical examination, chest pain, abdominal pain, or an abdominal mass as the main complaint. The prognosis was assessed by recurrence and/or metastasis after surgery, and the patients were divided into a good prognosis group and a poor prognosis group. The patients were categorized into four groups based on the percentage composition of fatty, myxoid, and non-fatty/non-myxoid elements: 0%, <25%, 25%–50%, and >50% (with the latter further divided into 50%–75% and >75% subgroups). The institutional review board approved this retrospective study and waived the requirement for informed consent. All patients underwent conventional postoperative chemoradiotherapy, including eight with recurrence and/or metastasis, 11 without recurrence and/or metastasis, and one who died during the follow-up (mean follow-up = 4.2 years; range = 0.5–8 years).

Approval was granted from the local ethical committee of the affiliated Jinhua Hospital, Zhejiang University School of Medicine, and the study was performed in accordance with the principles of the Declaration of Helsinki. Written informed consent was obtained from all patients, including the patient who died during the study.

MRI protocol

The MRI examinations were performed using two MRI scanners (Siemens MAGNETOM Avanto 1.5 T, Erlangen, Germany; GE Optima MR360 1.5 T, GE Healthcare, Milwaukee, WI, USA) with a standard protocol. An eight-channel phased-array body coil was used for imaging the chest, abdomen, and extremities. Subsequently, 0.1 mmol/kg body weight gadolinium-diethylenetriamine pentaacetic acid (Gd-DTPA) was administered, and fat-suppressed T1-weighted (T1W) sequences were acquired. The specific scanning protocol is provided in Table 1.

Table 1.

List of MRI acquisition protocols.

Acquisition site	Scan protocols		FOV (mm)	Section thickness/layer spacing (mm)	Matrix size
Abdomen/chest	T1W (breath-hold) T2W (breath-trigger)	Fat-suppressed	380 × 380	4/0.4	320 × 320
		Non-fat-suppressed
	DWI (breath-trigger)	b = 800
	Dual-phase (breath-hold)	In-phase
		Out-phase
	Contrast-enhanced T1W imaging (three phases)	Venous (axial plane)
		Arterial (axial and coronal plane)	380 × 380/400 × 400
		Delayed (axial plane)	380 × 380
Extremities	T1WI-FSE, T2WI-FSE	Fat-suppressed	340 × 255	Upper extremities 4/0.8 Lower extremities 5/0.8	360 × 256
		Non-fat-suppressed
	DWI	b = 1000
	Contrast-enhanced T1W imaging	Axial plane
		Coronal plane	420 × 420
		Sagittal plane

FSE, fast spin-echo; T1W, T1-weighted; T2W, T2-weighted.

MRI analysis

The MRI features and the component proportions of MLS were independently analyzed by two experienced blinded radiological physicians (JJ Hua and ML Ying) and a consensus was reached after discussion in cases of disagreement. Postoperative histopathological diagnosis is considered to be the gold standard. The imaging evaluation criteria included the parameters listed below.

Location and size of the tumor

The basic characteristics of the MLS lesions, including location, size, morphology, presence of capsules, necrotic or cystic changes, metastasis, invasion, calcification, and hemorrhage, were comprehensively analyzed. The tumor size was quantified by its maximum diameter in centimeters (cm) and volume in cubic centimeters (cm³), which was evaluated by segmentation of transverse contrast-enhanced T1W imaging. The tumor boundaries were defined manually, and the areas were automatically calculated using a workstation (AW4.6; GE Healthcare). The tumor volume was calculated according to the following equation:

$$\begin{array}{rl}\mathrm{V}=& \mathrm{S}\times (\mathrm{a}1+\mathrm{a}2+\mathrm{a}3\dots +\mathrm{an})\\ & (S:sectioninterval,{\mathrm{a}}_1\sim {\mathrm{a}}_{\mathrm{n}}:\mathrm{area}\mathrm{ofthesection})\end{array}$$

Tumor composition

On MRI scans, the components and imaging findings of the tumor were distinguished by visual estimation of T1W imaging, T2-weighted (T2W) imaging, diffusion-weighted (DWI), and T1W enhancement. There were three components: myxoid, fatty, and contrast-enhancing non-fatty/non-myxoid areas. T1W imaging of the myxoid area revealed low signal intensity (SI), whereas T2W imaging revealed very high SI with a fluid-like signal. Enhancement was either absent or presented as an extensive, patchy, or flocculent enhancement. However, non-fatty/non-myxoid areas presented high signals in T2W imaging (lower than those in myxoid areas) and low signals in T1W imaging, with significant enhancement. In the fat area, both T1W and T2W imaging exhibited high SI, while fat-suppressed imaging was low, and no enhancement was observed.

The volumes of fatty, myxoid, and non-fatty/non-myxoid tumor components were evaluated using the aforementioned segmentation technique. The mean percentage of each component relative to the total tumor volume was subsequently recorded, with a minimum increase of 5%. In addition, the degree of contrast enhancement of each component of the tumor was classified as no, nodular, heterogeneous, homogeneous, or diffuse. The region of interest (ROI) for measuring the apparent diffusion coefficient (ADC) values (×10⁻³ mm²/s) in the non-fat/non-myxoid area was ≥0.3 cm², and at least three measurements were recorded to obtain the mean value. Targeted regions with prominent intratumoral necrosis, cystic changes, hemorrhage, calcifications, artifacts, or partial volume effects were eliminated when the contours of the ROI were outlined.

Statistical analysis

All statistical analyses were conducted using SPSS version 22.0 (IBM Corp., Armonk, NY, USA). For continuous variables, normally distributed data were summarized as the mean ± SD, whereas skewed distributions were described as medians and interquartile range (IQR). Categorical variables were presented as frequencies and percentages. Group comparisons for normally distributed data were conducted using the independent samples t-test, whereas skewed data were analyzed using the Mann–Whitney U-test. For comparisons involving multiple groups, a one-way analysis of variance (ANOVA) was performed. Categorical data were analyzed using Fisher's exact test (due to the small sample size). A receiver operating characteristic (ROC) analysis was conducted to identify radiographic tumor components and better predict the prognosis of MLS patients. Two-tailed P values <0.05 were considered significant.

Results

Clinical characteristics and imaging findings

In total, 20 cases of MLS lesions were located in the retroperitoneum, deep muscles of the limbs, chest wall, posterior mediastinum, and pelvic and abdominal cavities. Patients were divided into a good prognosis group (n = 11) and a poor prognosis group (n = 9), with mean ages of 55.18 ± 18.34 years and 53.88 ± 8.96 years, respectively (P = 0.849) (Table 2). The morphology appeared elliptical, circular, or lobulated, with a mean maximum diameter of 13.34 ± 6.93 mm. In the good prognosis group, the tumor volume was 512.50 cm³ (IQR = 126.36–2237.20 cm³), whereas in the poor prognosis group, the corresponding value was 2417.40 cm³ (IQR = 296.91–4573.85 cm³) (Tables 2 and 3). Most of the lesions presented a low signal in T1W imaging and a heterogeneous high signal in T2W and DWI. Nine patients showed the presence of a capsule, whereas 11 were uncapsulated. Four patients presented with nearby tissue or organ involvement, three presented with accompanying cystic degeneration, and two presented with hemorrhage (Figs. 1 and 2).

Table 2.

Comparison of the MRI features in the two MLS groups ( $\overline{\chi }\pm \mathrm{s}$ , n, %, M [Q_U, Q_L]).

Characteristics	Good prognosis (n = 11)	Poor prognosis (n = 9)	t/Z	P
Age (years)	55.18 ± 18.34	53.88 ± 8.96	0.193	0.849 (NS)
MD (cm)*	10.80 (8.10–14.30)	18.50 (9.40–22.95)	−1.140	0.254 (NS)
Tumor volume (cm3)*	512.50 (126.36–2237.20)	2417.40 (296.91–4573.85)	−0.798	0.425 (NS)
NCD (%)	0	3/9 (33.33)	NA	NA
Hemorrhage (%)	0	2/9 (22.22)	NA	NA
Infiltration (%)	0	4/9 (44.44)	NA	NA
Tumor capsule	6/11 (54.54)	3/9 (33.33)	NA	0.650 (NS)
ADC value (×10–3mm2/s)	1.66 ± 0.23	1.21 ± 0.41	3.151	0.006
Tumor components (%)
ADC low signal*	5 (0.00–10.00)	20.0 (11.00–39.00)	−3.194	0.001
Non-fatty/non-myxoid*	15.00 (10.00–20.00)	70.00 (52.50–77.50)	−3.718	<0.001
Myxoid*	75.00 (65.00–85.00)	25.00 (17.50–42.50)	−3.583	<0.001
Fatty*	5 (0.00–15.00)	5 (0.00–7.50)	−0.702	0.483 (NS)

Values are given as n (%), mean ± SD, or median (IQR). The presence of tumor capsules was compared between the two groups using Fisher's exact test.

^*Does not comply with normal distribution.

ADC, apparent diffusion coefficient; MD, maximum diameter; MLS, myxoid liposarcoma; MRI, magnetic resonance imaging; NA, not applicable; NCD, necrosis and cyst degeneration; NS, not significant.

Table 3.

MRI features and prognosis of 20 patients with MLS.

Case	MD (cm)	Volume (cm3)	Fatty		NF/NM		Myxoid		ADC low signal (mm2/s)		Postoperative follow-up
			Fraction %	Shape	Fraction %	ENH	Fraction %	ENH	Fraction %	Value	Time (years)	Outcome
1	6.3	87.3	0	NA	15	HET	85	Diffuse	0	1.63	5	NRM
2	9.9	176.4	10	Divided	10	HET	80	Diffuse	0	1.95	5.5	NRM
3	14.3	2237.2	0	NA	15	HOM	85	Diffuse	10	1.11	7	NRM
4	9.6	257.2	0	NA	40	HOM	60	Diffuse	20	1.71	2.5	Recurrence
5	12.6	634.7	0	NA	45	HET	55	Diffuse	5	1.76	5	NRM
6	9.2	418.1	5	Divided	75	HET	20	ND	15	1.34	3	RM
7	22.4	5890.3	5	Nodular	80	HET	15	Diffuse	60	0.75	0.5	Death
8	24.7	3141.6	5	Divided	55	HET	40	Diffuse	10	1.92	1.5	Recurrence
9	3.2	8.1	0	NA	70	HET	30	Diffuse	5	0.85	4	Recurrence
10	5.6	43.0	0	NA	35	HET	65	Diffuse	0	1.88	6	NRM
11	20.1	3290.6	0	NA	50	HET	50	Diffuse	20	1.06	2.5	Recurrence
12	18.5	2417.4	10	Nodular	70	HET	20	ND	40	0.79	1	Recurrence
13	12.1	1007.3	45	Diffuse	10	HET	45	Diffuse	5	1.62	8	NRM
14	8.1	126.36	5	Divided	10	HET	85	Nodular	5	1.54	5	NRM
15	23.5	5857.1	0	NA	75	HET	25	Diffuse	50	1.12	2	RM
16	10.8	512.5	15	Divided	10	HET	75	Diffuse	15	1.78	5.5	NRM
17	9.6	336.6	0	NA	85	HET	15	Diffuse	50	1.35	3	Recurrence
18	22.4	7010.3	5	Divided	20	HET	75	ND	0	1.56	5	NRM
19	8.5	403.3	0	NA	20	HOM	80	Diffuse	0	1.83	6	NRM
20	18.4	4736.2	15	Divided	10	HET	75	Diffuse	10	1.65	5	NRM

ADC, apparent diffusion coefficient (mm²/s); ENH, enhancement; HET, heterogeneous; HOM, homogeneous; MD, maximum diameter; MLS, myxoid liposarcoma; MRI, magnetic resonance imaging; ND, nodular and diffuse; NF/NM, non-fatty/non-myxoid; NRM, No recurrence or metastasis; RM, recurrence and metastasis.

Fig. 1.

A 61-year-old male patient with MLS presented with abdominal pain for more than 1 month. (a) The oval mass with non-fatty/non-myxoid areas >50% located in the left retroperitoneum showed mixed SI on T2W imaging with fat suppression. (b) Diffuse and heterogeneous high SI on DWI. (c) Diffuse and heterogeneous low SI in the ADC (the lowest value was 1.26 × 10^–3/mm²). (d) Pronounced and heterogeneous enhancement in enhanced imaging, in which the flocculent enhanced or non-enhanced area was myxoid (T2W imaging revealed significant high signal areas) and the left kidney was invaded. (e) Under the microscope, the tumor cells were star-like, round or ovoid, and cytoplasmic eosinophilic with interstitial division and large amounts of myxoid and “chicken-wire” capillaries (arrows) (H&E staining ×20). (f) Recurrence occurred after 2.5 years of follow-up. A large, heterogeneous, enhancing mass was found in the left retroperitoneum. ADC, apparent diffusion coefficient; DWI, diffusion-weighted imaging; MLS, myxoid liposarcoma; SI, signal intensity; T2W, T2-weighted.

Fig. 2.

A 26-year-old woman was diagnosed with MLS. (a) On T1W imaging (sagittal) of the left forearm, an oval-shaped mass presented low SI, with an internal “marble textural pattern” stripe-shaped high S (fat) partition, and a clear boundary. (b) On T2W imaging, the mass showed high SI with low signal separation. (c, d) DWI and ADC showed high signals, with a mean ADC value of 2.29 × 10⁻³ mm²/s (myxoid areas). (e) T1W imaging enhancement (sagittal) revealed heterogeneous and obvious enhancement in non-fatty/non-myxoid areas (<25%), weak enhancement in myxoid areas (>75%), and no enhancement in fat areas (arrow). (f) Short spindle tumor cells, fat blasts, and rich capillaries were found under the microscope (H&E staining ×200). The mucus matrix was rich enough to form a mucus lake, and the marginal cells were compressed and flattened, similar to “pulmonary edema” (arrow). ADC, apparent diffusion coefficient; DWI, diffusion-weighted imaging; MLS, myxoid liposarcoma; SI, signal intensity; T1W, T1-weighted; T2W, T2-weighted.

Imaging characteristics of different tumor components

Non-fatty/non-myxoid components presented high SI on DWI and correspondingly low or moderately low SI on ADC maps. They manifested homogeneous or heterogeneous enhancement, whereas myxoid components presented no enhancement or nodular, patchy, or strand-like enhancement. Fatty components showed septal, nodular, or patchy hyperintensities on non-fat-suppressed T1W imaging. In one case, a “marble textural pattern” was found, which did not exhibit contrast enhancement. Further details are described in Table 3 and Figs. 1 and 2.

MLS component proportion analysis

All patients presented different percentages of non-fatty/non-myxoid and myxoid components. The average proportion of non-fatty/non-myxoid components in MRL lesions was 15.00% (IQR = 10.00%–20.00%) in the good prognosis group and 70.00% (IQR = 52.50%–77.50%) in the poor prognosis group (P < 0.001). In the good prognosis group, the mean ADC value of the non-fatty/non-myxoid region was 1.66 ± 0.23 mm²/s, with a low signal proportion of 5% (IQR = 0.00%–10.00%). In contrast, the corresponding values were 1.21 ± 0.41 mm²/s and 20.00% (IQR = 11.00%–39.00%) in the poor prognosis group (P = 0.006 and P = 0.001). In the <25% group, all nine patients had a good prognosis, whereas in the >50% group, all seven patients had a poor prognosis (P < 0.001). This included five cases in the 50%–75% subgroup and two cases in the >75% subgroup. In the 25%–50% group, there were two patients each with good and poor prognoses, with no significant differences compared to the other groups. In addition, in the former group, the proportion of myxoid components in the lesions was 75.00% (IQR = 65.00%–85.00%), whereas, in the poor prognosis group, it was 25.00% (IQR = 17.50%–42.50%; P < 0.001). The prognosis for the group with a mucus composition >50% was significantly better than that for the group with a mucus composition <25% (P = 0.005). Specifically, the >50% group included nine patients with a good prognosis, one patient with a poor prognosis, four patients with a good prognosis in the 50%–75% subgroup, and five patients with a good prognosis in the >75% subgroup. In contrast, four patients in the <25% group had a poor prognosis. In the 25%–50% group, four patients had a poor prognosis, and two patients had a good prognosis, which was significantly different from the >50% group (P = 0.036). The percentage of fat in the good prognosis group was 5% (IQR = 0.00%–15.00%), whereas that in the poor prognosis group was 5% (IQR = 0.00%–7.50%). The difference between the two groups was not significant (P = 0.483). Five patients with no fat content had a good prognosis, whereas the other five had a poor prognosis. Among the patients whose fat content was <25%, five patients had a good prognosis and four patients had a poor prognosis. One patient in the 25%–50% fat content group had a good prognosis. None of the patients had a fat content >50% (Tables 2 and 4).

Table 4.

Comparison of different proportions of MLS components and prognosis(n, %)

Fraction Grade		Fatty	Good/Poor	Non-fatty /non- myxoid	Good/Poor	Myxoid	Good/Poor
0%		10,50.0%	5/5	0	n.a.	0	n.a.
＜25%		9,45.0%	5/4	9,45.0%a	9/0	4,20.0%d	0/4
25-50%		1,5.0%	1/0	4,20.0%b	2/2	6,25.0%e	2/4
＞50%		0	n.a.	7,35.0%c	0/7	10,50.0%f	9/1
50-75%		0	n.a.	5,25.0%	0/5	5,26.7%	4/1
＞75%		0	n.a.	2,10.0%	0/2	5,25.0%	5/0
P		n.a.	0.944(n.s.)	n.a.	n.a.	n.a.	n.a.
Anova P		n.a.	n.a.	n.a.	＜0.001	n.a.	0.002

Good, Good prognosis; Poor, Poor prognosis.n.a:not applicable；n.s.:not significant; ^abP=0.077, ^acP＜0.001, ^bcP=0.467;^deP=0.103,^dfP=0.005,

^efP=0.036.a, d:＜25% grooup;b, e:25-50% group;c, f:＞50% group.

Prognostic evaluation efficacy of the tumor component proportion model

In this study, ROC curves were plotted to analyze the prognostic value of tumor component proportion models, including non-fatty/non-myxoid, myxoid, and fat components. The results revealed that the non-fatty/non-myxoid component proportion model achieved the highest diagnostic efficacy, with an area under the ROC curve (AUC) of 0.99, sensitivity of 100%, specificity of 91.17%, negative predictive value (NPV) of 100%, positive predictive value (PPV) of 88.89%, accuracy of 88.89%, and cutoff value of 37.50%. In the myxoid component model, the AUC was 0.98, sensitivity was 80.00%, accuracy and PPV reached 88.89%, and the specificity and NPV were 90.00%, 81.82%, and and cutoff value of 60.00% respectively. In the ADC low-signaling component model, the AUC was 0.93, with a sensitivity of 77.78%, specificity of 90.09%, and NPV of 83.33%. The accuracy and PPV were 87.50%, with a cutoff value of 12.50%. In contrast, the indices of the fat component model were relatively low (Table 5, Fig. 3).

Table 5.

Fatty, myxoid, non-fatty/non-myxoid and ADC low signal diagnose efficiency of MLS with poor prognosis.

Model	AUC	SEN (%)	SPE (%)	ACC (%)	PPV (%)	NPV (%)	Cutoff values (%)
Fatty	0.61	11.11	90.09	50.00	50.00	55.56	15.00
Myxoid	0.98	80.00	90.00	88.89	88.89	81.82	60.00
Non-fatty/non-myxoid	0.99	100.00	91.17	88.89	88.89	100.00	37.50
ADC low signal	0.93	77.78	90.09	87.50	87.50	83.33	12.50

ACC, accuracy; ADC, apparent diffusion coefficient; AUC, area under the receiver operating characteristic curve; MLS, myxoid liposarcoma; NPV, negative predictive value; PPV, positive predictive value; SEN, sensitivity; SPE, specificity.

Fig. 3.

The dotted lines in different colors indicate that the ROC curve for each model corresponds to a different AUC, which represents the accuracy of diagnosis in different models. The dotted lines in different colors indicate the ROC curves of the fatty, myxoid non-fatty/non-myxoid, and low ADC signals. ADC, apparent diffusion coefficient; AUC, area under the ROC curve; ROC, receiver operating characteristic.

Discussion

As shown in other studies (7,10,17), MLS may occur in the deep muscles of limbs, with a median age of approximately 50 years, which is consistent with our findings. MLS is a multinodular tumor with low to moderate cellular architecture and a prominent myxoid matrix similar to “pulmonary edema,” rich in hyaluronic acid, and a characteristic “chicken claw-like” capillary plexus (1,17,18). The MR features of MLS are determined based on their histological components, which appear as hyperintensities on T2W imaging. Most incidences of MLS show a clear boundary, and some are wrapped. The enhancement may be homogeneous or heterogeneous. “Marble flower pattern” adipose tissue is present in masses with general hyperintensity on T1W imaging, especially in MLS (17,19–21). However, 57.9% of MLS contains no fat, and the content usually accounts for <10% (7).

In this study, 10/20 patients contained fat components; in most cases, the percentage of fat was <10%, which was consistent with previous reports. MLS grows slowly, is painless, and occurs deep enough such that when found, the tumor is often large, with a median size of about 12 cm (22), and 6% of patients are metastatic at the time of first diagnosis (6,16). Gimber et al. (7) reported a strong correlation between tumors >10 cm and high-grade MLS. However, Shinoda et al. (23) suggested that the survival rate of MLS was not related to the size of the tumor. This study revealed no significant difference in the average maximum diameter of the lesions between the good and poor prognosis groups, resulting in inconsistent results. These differences may be related to the differences in sample size or treatment regimens. However, we speculate that the malignancy of tumors is primarily determined by their composition and heterogeneity, particularly the presence of round cell components.

The proportion of round cell components in MLS is directly related to patient prognosis (24). Pure myxoid MLS is considered to be a low-grade sarcoma with slow growth and a long survival time (14). However, MLS with >5% round cell component is regarded as a high-grade sarcoma with high recurrence, metastasis, and poor prognosis (14). Lemeur and Haniball et al. (6,25,26) reported that MLS with a round cell composition of more than 5% had a greater risk of recurrence, metastasis, and death, and the local recurrence rate reached 3.86 times. Lansu et al. (24) reported that for localized tumors, the 5-year overall survival rate was 78% and the 10-year overall survival rate was 66%. In comparison, for patients who presented with metastatic disease, overall survival was 47% at 1 year and 24% at 3 years, with a median overall survival of 10 months. Studies have found that neoadjuvant chemotherapy and radiotherapy can reduce the possibility of recurrence and metastasis in high-grade tumors (10). Tumors are diagnosed and graded by needle biopsy before surgery. However, the tumor grade risk is underestimated when the high-grade component (the round cell component) is not sampled (7). Therefore, analyzing the tumor components based on imaging features may be valuable for determining the proportion of round cells within the tumor.

A study conducted by Lowenthal et al. (14) revealed that high-grade MLS had a significantly greater mean proportion of non-fatty/non-myxoid areas than low-grade MLS (50 ± 25% vs. 2 ± 9%; P < 0.001). The mean proportion of myxoid areas in the low-grade MLS was approximately 88% ± 16%, whereas that in the high-grade MLS was only 45% ± 25% (P < 0.001). However, the mean proportion of fat in low-grade MLS was 10.00% ± 11.00%, whereas it was 6.00% ± 4.00% in high-grade MLS (P = 0.66). A retrospective study by Tobajas et al. (15) revealed that tumors exhibiting a myxoid component <25% were classified as high-grade (P = 0.01). In addition, 83.3% of the tumors with a non-fatty, non-myxoid component exceeding 50% were also categorized as high-grade (P = 0.03). Moreover, 61.5% of these tumors contained >5% round cells, which was significant (P = 0.01). In our study, the proportion of non-fatty/non-myxoid areas was lower in the good prognosis group (15.00%, IQR = 10.00%–20.00%) than in the poor prognosis group (70.00%, IQR = 52.50%–77.50%; P < 0.001). In contrast, compared to 25.00% (IQR = 17.50%–42.50%) in the poor prognosis group, the proportion of the myxoid area was greater in the good prognosis group (75.00%, IQR = 65.00%–85.00%; P = 0.003). The median proportion of fat was not significantly different between the two groups (5% in both the poor prognosis group and the good prognosis group) (P = 0.483). The above results were similar to those reported in previous studies. Tobajas et al. (15) reported that high-grade MLS had non-fatty/non-myxoid components >75% or myxoid components <25%. In addition, >50% non-fatty/non-myxoid components resulted in a seven-fold greater risk of high-grade MLS and a four-fold greater risk of round cell content. In this study, those with a non-fatty/non-myxoid component ratio of <25% had a good prognosis, whereas those with a non-myxoid component ratio of >50% had a poor prognosis. The group with a myxoid composition of <25% had a poor prognosis, whereas the >50% group had a good prognosis in nine patients and a poor prognosis in one patient, which was consistent with the results reported in previous studies.

Compared to the good prognosis group, the poor prognosis group had a lower mean ADC value and a greater proportion of low ADC signals in the non-fatty/non-myxoid areas (P = 0.006 and P = 0.001). However, Tobajas et al. (15) reported the opposite result. This difference occurred probably because the ADC values and proportion of low ADC values, including those in the myxoid area, were measured; the diffusion-restricted area of MLS is mainly concentrated in the non-fatty/non-myxoid area, resulting in a greater proportion of low ADC values in this area, whereas the myxoid area of tumors had a higher ADC value and a lower proportion of low ADC values. Therefore, we hypothesized that a greater proportion of non-fatty/non-myxoid components may predict high-grade MLS, which may histologically resemble round cell clusters, especially areas with a high proportion of low ADC signals.

The present study has some limitations. We conducted a retrospective study. Although the small sample size introduced some bias, it also reflected the low incidence rate of this disease. Future studies with larger sample sizes need to be performed to confirm our findings. In addition, while the absence of a comparison with histological round cell components is a significant limitation, our study, when considered along with previous literature, indirectly supported some of their ideas.

In conclusion, the MRI features of MLS were directly correlated with histology. Regions of restricted diffusion in the non-fatty/non-myxoid component may indicate the presence of a dense round cell component. MRI features of different MLS components are promising tools for predicting prognosis and provide a reference for conducting clinical diagnosis and making treatment decisions, especially before performing preoperative punch biopsy.