MRI analysis of undifferentiated pleomorphic sarcoma: correlating imaging features with histological grade

Introduction

Undifferentiated pleomorphic sarcoma (UPS), previously called malignant fibrous histiocytoma, is the most common histological subtype of soft tissue sarcoma (STS), accounting for approximately 20% of all cases (1-3). It is a diagnosis of exclusion, defined as a high-grade mesenchymal tumor lacking specific lines of differentiation on histopathology (4).

UPS typically presents as a rapidly enlarging deep soft tissue mass with aggressive behavior and poor prognosis (5). Histological grading, particularly using the Fédération Nationale des Centres de Lutte Contre le Cancer (FNCLCC) system, is critical for treatment planning and prognostic assessment (6). However, accurate preoperative grading remains challenging due to pronounced intratumoral heterogeneity, and core needle biopsy may underestimate tumor grade (7). This limitation highlights the urgent need for non-invasive methods capable of capturing tumor heterogeneity and predicting histological grade reliably across the entire tumor volume.

Magnetic resonance imaging (MRI) enables non-invasive assessment of tumor characteristics, including necrosis, cellularity, and vascularity. These features, derived from conventional sequences and diffusion-weighted imaging (DWI), may reflect tumor aggressiveness and histological grade (8,9). In recent years, studies have increasingly explored the use of MRI—including radiomics approaches—to assess tumor heterogeneity and predict histological grade in STSs (10,11). However, these investigations have largely focused on STSs as a collective group. Given the low incidence of UPS, studies focusing specifically on this biologically aggressive and histologically heterogeneous subtype remain scarce.

This study aimed to investigate whether MRI-based imaging features, including both conventional and DWI parameters, can serve as non-invasive biomarkers for assessing the histological grade of UPS. Furthermore, we aimed to evaluate the association between imaging characteristics and pathological grading, as well as their potential clinical relevance in improving diagnostic accuracy and guiding individualized treatment strategies. We present this article in accordance with the STARD reporting checklist (available at https://qims.amegroups.com/article/view/10.21037/qims-2025-1-2778/rc).

Methods

The study was conducted in accordance with the Declaration of Helsinki and its subsequent amendments. This study was approved by the Ethics Committee of Henan Cancer Hospital (approval No. 2023-KY-0063-001). The waiver of informed consent was granted because the study was retrospective in nature and did not involve any additional interventions or procedures beyond routine clinical care. A total of 106 patients with pathologically-confirmed UPS were retrospectively enrolled from three sarcoma referral centers between January 2015 and December 2023. The diagnosis of UPS was established according to the World Health Organization (WHO) classification as a diagnosis of exclusion, based on histomorphological features showing high-grade pleomorphic sarcoma without identifiable lines of differentiation. Tumors with specific differentiation, such as myxofibrosarcoma or dedifferentiated liposarcoma, were excluded following detailed pathological evaluation by experienced sarcoma pathologists. Patients were included if they met all of the following conditions: histological diagnosis of UPS based on postoperative-confirmed pathological evaluation; no prior oncologic treatments administered before MRI examination, including surgery, chemotherapy, radiotherapy, immunotherapy, or targeted therapy; availability of high-quality pre-treatment MRI, including T1-weighted, T2-weighted, and DWI, acquired within four weeks prior to treatment initiation; complete clinical, radiological, and pathological records accessible for retrospective analysis; surgery involved wide resection (R0) and all patients received radiotherapy/chemotherapy. Patients were excluded under the following conditions: history of other malignancies or co-existing primary tumors at diagnosis; MRI contraindications, such as pacemaker implantation or documented severe contrast agent allergy; Prior exposure to oncologic therapies likely to alter tumor morphology or imaging features before MRI acquisition; Incomplete imaging datasets or insufficient histopathological information to determine tumor grade (Figure 1). Overall survival (OS) was defined as the interval from curative surgery to the last follow-up (censored data) or death (irrespective of the cause).

Figure 1 Case selection process flowchart. MRI, magnetic resonance imaging; UPS, undifferentiated pleomorphic sarcoma.

MRI protocol

All patients underwent pre-treatment MRI examinations using 3.0 Tesla (3T) MRI systems at the participating centers, including a Philips Ingenia 3.0T scanner (Philips, Amsterdam, the Netherlands) and a Siemens MAGNETOM Skyra 3.0T scanner (Siemens, Erlangen, Germany). Both systems performed standardized imaging protocols comprising conventional anatomical sequences, and DWI sequences. Minor variations in imaging parameters were present across devices due to hardware specifications.

For the Philips Ingenia 3.0T system, conventional sequences included: T1-weighted imaging (T1WI) using turbo spin echo with repetition time/echo time (TR/TE) =400 ms/10 ms, matrix size: 380×260, field of view (FoV): 380×260 mm, number of averages (NEX): 1, acquisition time 1 minute 27 seconds; T2-weighted imaging (T2WI) using turbo spin echo with TR/TE =3,500 ms/90 ms, matrix size: 380×260, FoV: 380×260 mm, NEX: 1, acquisition time 1 minute 48 seconds; T2-weighted fat-suppressed (T2WI-FS) with TR/TE =3,500 ms/90 ms, matrix size: 380×260, FoV: 380×260 mm, NEX: 1, acquisition time 1 minute 48 seconds; DWI (echo planar imaging, TR/TE =3,800 ms/65 ms) with b-values of 0 and 800 s/mm², matrix size: 136×104, FoV: 380×260 mm, NEX per b-value: 2, acquisition time 2 minutes 15 seconds; contrast-enhanced T1WI (CE-T1WI) using turbo spin echo with TR/TE =5.1 ms/2.3 ms, matrix size: 380×260, FoV: 380×260 mm, acquisition time 2 minutes 30 seconds. A gadolinium-based contrast agent was injected intravenously via the antecubital vein at a dose of 0.1 mmol/kg body weight, using a high-pressure injector at a flow rate of 2.5 mL/s, followed by a saline flush.

For the Siemens MAGNETOM Skyra 3.0T system, imaging sequences included: T1WI with turbo spin echo with TR/TE =400 ms/10 ms, matrix size: 384×288, FoV: 380×260 mm, NEX: 1, acquisition time 1 minute 35 seconds; T2WI with turbo spin echo with TR/TE =3,500 ms/90 ms, matrix size: 384×288, FoV: 380×260 mm, NEX: 1, acquisition time 1 minute 54 seconds; T2WI-FS with TR/TE =3,500 ms/90 ms, matrix size: 384×288, FoV: 380×260 mm, NEX: 1, acquisition time 1 minute 54 seconds; DWI (echo planar imaging) performed with TR/TE =4,000 ms/67 ms and b-values of 50 and 800 s/mm², matrix size: 128×128, FoV: 380×260 mm, NEX per b-value: 2, acquisition time 2 minutes 20 seconds; CE-T1WI with TR/TE =5.3 ms/2.3 ms, matrix size: 384×288, FoV: 380×260 mm, acquisition time 2 minutes 30 seconds; Gadodiamide (GE Healthcare, Chicago, IL, USA) was administered as the contrast agent at a dosage of 0.1 mmol/kg, delivered via a power injector at a rate of 2.5 mL/s, followed by a 20 mL saline flush to ensure consistent bolus delivery. However, NEX for the b-value of 800 s/mm² was set to 2 for the DWI sequence to balance scan time with adequate signal-to-noise ratio (SNR) in our cohort. Although NEX is relatively low, the imaging protocol used parallel imaging techniques (e.g., SENSE for Philips and GRAPPA for Siemens scanners) to improve SNR efficiency while maintaining acceptable spatial resolution.

All MRI features of UPS cases were independently reviewed in a double-blinded fashion by three radiologists (C.L., F.M., and C.X., with 6, 10, and 15 years of experience in sarcoma imaging, respectively), using a standardized assessment framework. The evaluation encompassed both qualitative and quantitative parameters, as follows (12): (I) growth pattern was assessed based on T2WI and contrast-enhanced sequences, and categorized into three types according to tumor margins and infiltration characteristics: pushing, focal infiltrative, and diffuse infiltrative; (II) signal intensity (SI) heterogeneity on conventional T1WI and T2WI, visually classified as <50% or ≥50% heterogeneous signal distribution within the tumor. Heterogeneity was assessed based on the visual distribution of SI variation within the lesion, including areas suggestive of necrosis, hemorrhage, cystic degeneration, or internal structural complexity. A predefined semi-quantitative threshold was applied to classify heterogeneity into two categories: <50% and ≥50% of the tumor volume. Lesions with relatively uniform SI were categorized as <50% heterogeneity, whereas those demonstrating visually appreciable and extensive signal variation involving at least half of the tumor were categorized as ≥50% heterogeneity; (III) post-contrast SI heterogeneity on T1WI following gadolinium-based contrast administration, similarly stratified as <50% or ≥50% heterogeneous signal area; (IV) tumor necrosis, defined radiologically as non-enhancing regions on contrast-enhanced T1WI that appear hyperintense on T2WI, with necrotic volume proportion categorized as <50% or ≥50% of the total tumor volume; (V) peritumoral edema, identified as high fluid-like signal on T2WI extending beyond the solid tumor margin, and distinguishable from tumor infiltration; (VI) peritumoral enhancement, defined as abnormal contrast uptake extending beyond the clearly defined tumor border on enhanced T1WI sequences.

For apparent diffusion coefficient (ADC) analysis, tumor segmentation and quantitative heterogeneity assessment were performed using the open-source platform 3D Slicer (https://www.slicer.org/). The tumor boundary was manually delineated on each slice of the ADC map, with reference to corresponding T2WI and contrast-enhanced sequences, to ensure accurate localization. Regions of interest (ROIs) were drawn to encompass the entire solid tumor, explicitly excluding areas of cystic degeneration or necrosis—defined by high signal on T2WI, low signal on DWI, and lack of enhancement on post-contrast images. From the delineated three-dimensional (3D) volume of interest (VOI), a set of ADC-based heterogeneity metrics was extracted, including minimum, maximum, mean, and difference (defined as the difference between maximum and minimum ADC values) (13). Final measurements were obtained by averaging their results to minimize observer bias.

Histopathological assessment

All tumor specimens underwent histopathological evaluation by two independent pathologists (Y.W. and T.C., with 7 and 9 years of experience in sarcoma pathology, respectively), utilizing a double-blind review process. Each case was graded according to the standardized FNCLCC system, which incorporates criteria including tumor differentiation, mitotic activity, and extent of necrosis. Discrepancies in grading between the two observers were resolved through joint re-evaluation and consensus.

For subsequent analysis, tumors were stratified into two prognostic categories based on FNCLCC scores and supported by current evidence from neoadjuvant clinical studies: Grade III lesions were classified as high-grade sarcomas, whereas Grades I and II were collectively designated as low-grade tumors.

Statistical analysis

All statistical analyses were performed using the software SPSS 24.0 (IBM Corp., Armonk, NY, USA), GraphPad Prism version 9.0 (GraphPad Software, San Diego, CA, USA), and MedCalc version 20.1 (MedCalc Software, Ostend, Belgium). Continuous variables were assessed for normality using the Shapiro-Wilk test. Data conforming to a normal distribution were expressed as mean ± standard deviation (SD), whereas non-normally distributed variables were presented as median with interquartile range (IQR). Independent samples t-tests were used for normally distributed continuous variables, and the Mann-Whitney U test was applied for non-normally distributed data. Chi-squared or Fisher’s exact tests were used to analyze associations between categorical variables and tumor grade. Intraclass correlation coefficients (ICCs) and Fleiss kappa coefficient were calculated to assess interobserver agreement for qualitative and quantitative MRI features, with ICC/kappa values interpreted as follows: <0.40 (poor), 0.40–0.59 (fair), 0.60–0.74 (good), and ≥0.75 (excellent) consistency. Binary logistic regression was performed to identify imaging predictors of high-grade UPS. Variables with P<0.05 in univariate analysis were entered into a multivariate logistic model, and results were expressed as odds ratios (ORs) with 95% confidence intervals (CIs). For MRI features with statistically significant associations, receiver operating characteristic (ROC) curve analysis was conducted to assess diagnostic performance. The area under the curve (AUC), along with sensitivity and specificity, was reported to evaluate discriminatory capability. OS curves were generated using the Kaplan-Meier method and compared using the log-rank test. To assess the robustness and generalizability of the predictive model, we performed 5-fold cross-validation. A two-sided P value <0.05 was considered statistically significant in all analyses.

Results

Patient characteristics and distant metastasis outcomes

Among the 83 patients included in the study (mean age, 59.5±8.9 years; 45 men), 43 patients (51.8%) were younger than 60 years, whereas 40 patients (48.2%) were 60 years or older. The proportion of patients aged ≥60 years slightly increased with tumor grade, reaching 53.6% in the Grade III group. There were 6 patients (7.2%) with Grade I tumors, 21 (25.3%) with Grade II, and 56 (67.5%) with Grade III tumors. Anatomically, the lower limb was the most common tumor site, observed in 47 cases (56.6%), with the highest proportion in the Grade III group (60.7%).

Regarding tumor depth, deep lesions were more common in high-grade tumors (58.3% in Grade Ⅱ–III), whereas superficial lesions were more frequent in low-grade tumors (50.0% in Grade I). As for tumor size, lesions ≥10 cm were predominantly seen in the Grade III group (44.6%), whereas tumors <5 cm were most common in the Grade I group (83.3%).

A total of 30 patients (36.1%) had lymph node metastasis (N1), with the highest proportion observed in the Grade III group (42.9%). Distant metastasis (M1) was present in 23 patients (27.7%), again more frequent in the Grade III group (32.1%). Regarding Ki-67 proliferative index, 39 patients (47.0%) had a Ki-67 index ≥30%, most commonly in the Grade III group (51.8%), whereas only 1 patient (16.7%) in the Grade I group showed a Ki-67 index ≥30% (Table 1).

Relationship between MRI features and histologic grade

In terms of MRI characteristics, UPSs of different histologic grades exhibited significant differences in several features (Table 2). Regarding growth patterns, low-grade tumors were more likely to show a pushing type growth (44.5%), whereas high-grade tumors more frequently demonstrated a diffuse infiltrative pattern (50.0%), with a statistically significant difference (P=0.047).

On T2WI, high-grade tumors more commonly exhibited heterogeneous SI (≥50% heterogeneity: 64.3% vs. 44.7%), which was statistically significant (P=0.04). However, no significant differences were observed between the groups in terms of heterogeneity on T1WI or after gadolinium contrast enhancement (P=0.75 and 0.12, respectively). The proportion of tumor volume showing MRI signal consistent with necrosis was significantly higher in high-grade tumors (≥50% necrosis: 62.5% vs. 37.0%, P=0.03). Other imaging features, such as the degree of peritumoral edema (P=0.78) and presence of peritumoral enhancement (P=0.25), did not show statistically significant differences between the groups.

In DWI, the mean ADC value was significantly higher in low-grade tumors [(1.27±0.20)×10⁻³ mm²/s] compared to high-grade tumors [(1.10±0.21)×10⁻³ mm²/s, P=0.001]. Additionally, ADC difference (maximum ADC minus minimum ADC) was significantly greater in high-grade tumors [(1.53±0.30)×10⁻³vs. (1.24±0.27)×10⁻³ mm²/s, P<0.001], indicating greater internal heterogeneity in higher-grade lesions.

Assessing MRI features associated with high-grade UPS

Univariable logistic regression analysis identified several MRI features significantly associated with high-grade UPS (Table 3). These included growth pattern (OR, 2.25; 95% CI: 1.24–4.09; P=0.008), heterogeneous SI on T2-weighted imaging (OR, 2.62; 95% CI: 1.02–6.72; P=0.045), tumor volume with MRI signal compatible with necrosis (OR, 2.83; 95% CI: 1.10–7.33; P=0.032), lower ADC mean values (OR, 0.01; 95% CI: 0.001–0.20; P=0.002), and greater ADC difference (OR, 59.9; 95% CI: 6.22–577.23; P<0.001).

Multivariable logistic regression analysis confirmed that three variables remained independent predictors of high-grade tumors: growth pattern (OR, 2.51; 95% CI: 1.15–4.63; P=0.022); tumor volume with necrosis-like signal (OR, 4.53; 95% CI: 1.25–16.47; P=0.022); ADC difference (OR, 55.7; 95% CI: 4.02–773.91; P=0.003). In addition, ADC mean remained an independent inverse predictor (OR, 0.014; 95% CI: 0.001–0.63; P=0.028), indicating that lower ADC values are significantly associated with higher histologic grade (Figures 2,3).

Figure 2 A 65-year-old female with an irregular mass in the left thigh muscle. Postoperative pathological examination confirmed the diagnosis of low-grade UPS. (A) Axial T1-weighted imaging shows a well-defined mass within the quadriceps muscle, exhibiting iso- to mildly hypointense signal intensity relative to skeletal muscle. (B) Axial T2-weighted fat-suppressed imaging demonstrates predominantly high SI with mild internal heterogeneity involving <50% of the tumor volume. (C) DWI shows hyperintensity of the lesion. (D) The corresponding apparent diffusion coefficient map demonstrates moderate to mildly reduced signal intensity. (E,F) Axial and coronal contrast-enhanced T1-weighted fat-suppressed images demonstrate mild-to-moderate heterogeneous enhancement without extensive non-enhancing necrotic areas. (G) Immunohistochemistry Ki-67 immunohistochemical staining (×200) showed scattered positive nuclear cells, with a low Ki-67 proliferation index, suggesting that this undifferentiated pleomorphic sarcoma is actively growing but relatively localized. Follow-up at 58 months showed no recurrence or metastasis. DWI, diffusion-weighted imaging; SI, signal intensity; UPS, undifferentiated pleomorphic sarcoma.

Figure 3 A 57-year-old male with an irregular mass in the right thigh. Postoperative pathological examination confirmed the diagnosis of high-grade UPS. (A) Axial T1-weighted imaging shows a large intramuscular mass with iso- to slightly hypointense SI and ill-defined margins. (B) T2-weighted fat-suppressed imaging shows markedly heterogeneous high signal intensity, involving ≥50% of the tumor area. (C) Diffusion-weighted imaging showed most of the mass as high SI. (D) Apparent diffusion coefficient maps showed the corresponding region as low SI, consistent with the findings on DWI. (E,F) Axial and coronal contrast-enhanced T1-weighted fat-suppressed images demonstrate a large intramuscular mass with iso- to slightly hypointense SI and ill-defined margins and extensive heterogeneous enhancement. (G) Immunohistochemical Ki-67 staining (×200) showed that a large number of tumor cell nuclei are brown-positive, with a high Ki-67 index, suggesting that the tumor has active proliferative capacity. The patient died during follow-up at the 24th month. DWI, diffusion-weighted imaging; SI, signal intensity; UPS, undifferentiated pleomorphic sarcoma.

Diagnostic performance of MRI features for assessing high-grade UPS

ROC analysis was performed to evaluate the diagnostic performance of individual and combined MRI features in assessing high-grade UPS. Among the single parameters, ADC difference demonstrated the highest diagnostic performance, with an AUC of 0.755 (95% CI: 0.648–0.842), sensitivity of 91.07%, specificity of 48.15%, and a Youden index of 0.392 at a threshold of 1.186×10⁻³ mm²/s.

ADC mean also showed good discriminative ability, with an AUC of 0.700 (95% CI: 0.590–0.796), sensitivity of 67.86%, specificity of 66.67%, and a Youden index of 0.345 at a threshold of 1.179×10⁻³ mm²/s. Growth pattern and tumor volume with necrosis-related signal yielded lower AUC values (0.674 and 0.627, respectively), with corresponding Youden indices of 0.278 and 0.255. Notably, the combination of all assessing MRI parameters significantly improved diagnostic performance, yielding an AUC of 0.876 (95% CI: 0.786–0.938), with a sensitivity of 82.14% and specificity of 85.19%, and the highest overall Youden index of 0.673 (Table 4 and Figure 4). 5-fold cross-validation was performed to assess the predictive model’s robustness. The average AUC from cross-validation was 0.851, demonstrating that the model remains stable and generalizable across different subsets of the data.

Figure 4 Receiver operating characteristic analysis of different MRI features and combined model with high-grade UPS. ADC, apparent diffusion coefficient; MRI, magnetic resonance imaging; UPS, undifferentiated pleomorphic sarcoma.

Relationship between MRI features and OS in UPS patients

The follow-up period for patients ranged from 12 to 60 months (42.8±9.7 months). Among the 83 UPS patients, a total of 17 deaths were recorded. The results of the survival analysis are presented in Table 5 and Figure 5. By integrating MRI features independently associated with high-grade UPS, the following were identified: Growth pattern, specifically diffuse infiltrative (Chi-squared =26.76; P<0.001), tumor volume with MRI signal compatible with necrosis ≥50% (Chi-squared =6.73; P=0.009), and ADC mean <1.179×10⁻³ mm²/s (Chi-squared =6.57; P=0.010) were associated with poorer 5-year OS.

Figure 5 Kaplan-Meier curves for OS in 83 patients with UPS based on different MRI features. (A) Growth pattern: group 0 = pushing type, group 1 = focal infiltrative type, group 2 = diffuse infiltrative type. (B) Necrosis: group 0 = necrotic area <50%, group 1 = necrotic area ≥50%. (C) ADCmean: group 0 = ADCmean ≥1.179×10⁻³ mm²/s, group 1 = ADCmean <1.179×10⁻³ mm²/s. ADC, apparent diffusion coefficient; MRI, magnetic resonance imaging; OS, overall survival; UPS, undifferentiated pleomorphic sarcoma.

Interobserver agreement for MRI measurements

The ICC/kappa values for all MRI features were in the excellent range, ranging from 0.77 to 0.89.

Assessment of scanner-specific effects

Of the 83 patients included in the study, 42 were scanned using the Philips Ingenia 3.0T (Grade I: 2, Grade II: 13, Grade III: 27) and 41 using the Siemens MAGNETOM Skyra 3.0T (Grade I: 4, Grade II: 8, Grade III: 29). Subgroup analysis revealed no statistically significant differences in ADC values between the two scanners for either low-grade (P=0.14) or high-grade (P=0.46) tumors. Furthermore, multivariate regression analysis incorporating scanner type as a covariate confirmed that tumor grade remained an independent and significant predictor of ADC values (P<0.05), demonstrating that the primary findings were not confounded by scanner-specific parameterizations.

Discussion

The findings of this multicenter study further underscore the value of MRI in assessing the histological grade of UPS. By integrating conventional and functional MRI parameters—including tumor growth pattern, T2WI signal heterogeneity, necrotic proportion, and ADC metrics—we demonstrated significant differences between low- and high-grade UPS. These results are consistent with previous studies and highlight the ability of MRI to reflect tumor microarchitecture, supporting its role in radiologic grading of STSs (8).

Regarding growth behavior, low-grade UPS more commonly exhibited a “pushing-type” pattern with well-defined margins and clear separation from adjacent tissues. In contrast, high-grade tumors frequently showed diffuse infiltrative growth with ill-defined borders, indicating more aggressive biological behavior (14). This observation corroborates the findings of Chhabra et al. (8), who reported an association between poorly defined margins and higher histologic grade. The combination of qualitative assessment and a margin scoring system in our study improves objectivity and clinical applicability.

In terms of internal tumor characteristics, T2WI signal heterogeneity and necrotic proportion were key indicators of tumor biology. Tumors with ≥50% heterogeneous signal were predominantly high-grade and often demonstrated irregular high-intensity regions corresponding to necrosis, hemorrhage, or cystic change. Necrotic areas—typically non-enhancing on contrast-enhanced imaging and hyperintense on T2WI—were more frequently observed in high-grade UPS. Importantly, patients with diffuse infiltrative growth, necrotic signal ≥50%, and lower mean ADC values showed worse OS. These findings are consistent with those of Crombé et al. (9), supporting the association between heterogeneity, necrosis, and tumor grade. The quantitative evaluation applied in our study further strengthens the reproducibility of these imaging-pathology correlations.

From a functional perspective, mean ADC values were lower in high-grade tumors, whereas ADC variability was greater. Histologically, this reflects increased cellular density and reduced extracellular space, which restrict water diffusion and decrease ADC values (15). Structural complexity within high-grade tumors likely contributes to increased ADC heterogeneity. These findings are consistent with prior studies linking lower ADC values to higher malignancy (16). Importantly, incorporating ADC variability provides additional information beyond mean ADC alone, enabling a more comprehensive assessment of tumor microstructure. Multivariate analysis identified growth pattern, necrotic proportion, and ADC variability as independent predictors of high-grade UPS, whereas mean ADC was a negative predictor. The combined model improved diagnostic performance compared with single-parameter approaches (17).

In terms of diagnostic performance, ADC variability alone demonstrated good discriminative ability, whereas the combination of multiple imaging features further improved accuracy. This supports a multiparametric approach, which may serve as a practical radiological decision-support tool in clinical settings (12). Our findings are consistent with those of Hauwanga et al. (18), emphasizing the value of multi-modal MRI in improving grading accuracy.

The innovation of this study lies not only in the strategic selection and integration of imaging parameters but also in bridging conventional morphological interpretation with functional imaging analysis. By focusing on growth pattern, necrotic volume, and ADC variability, we introduce a novel, quantitative MRI-based assessing model that demonstrates independent value in differentiating UPS grades (9). This framework offers a reproducible and scalable basis for future research on standardized radiologic grading. Additionally, we explored the associations between imaging features and clinical variables such as patient age, tumor location, and lymphatic metastasis, further emphasizing the utility of MRI in comprehensive clinical management (15). Clinically, these findings hold significant implications. First, MRI emerges as a non-invasive and reproducible preoperative tool for histological grade assessment, which is particularly valuable when biopsy specimens are limited or subject to sampling bias (19). Second, the proposed multi-parameter model supports individualized treatment planning, enabling clinicians to better stratify patients for surgery, preoperative therapy, or tailored follow-up, thereby advancing the practice of precision medicine in STS management. We acknowledge that the choice of different lower b-values (0 vs. 50 s/mm²) across scanners may influence the calculated ADC values. As is known from the intravoxel incoherent motion model, ADC values calculated with b=0 s/mm² contain a higher fraction of microcapillary perfusion compared to those calculated with b=50 s/mm², where the perfusion effect is effectively suppressed. This heterogeneous perfusion weighting could theoretically act as a confounder. However, we have mitigated this effect by statistically adjusting for scanner type as a covariate in our analyses. Our results showed that the primary findings regarding tumor grading remained robust, indicating that the true diffusion restriction at high b-values (800 s/mm²) dominated the diagnostic value in our cohort. Nevertheless, although the contribution of microperfusion is reduced at b=50 s/mm², it is not fully suppressed and may still have a minor impact. Furthermore, although the 50% threshold for visual categorization of heterogeneity is somewhat subjective, it was adopted based on established literature to maximize inter-observer reproducibility in a clinical setting (9,18,20), prioritizing practical utility over complex computational segmentations. In future work, we plan to explore the effects of varying thresholds (e.g., 30%, 60%, 70%) on segmentation accuracy and their impact on MRI feature quantification. This will help to assess the reliability and generalizability of the proposed segmentation method in different tumor types and clinical settings. Additionally, we opted against using advanced texture-based analysis, as higher-order radiomic features are highly sensitive to the multi-scanner parameter variations present in our cohort (21-24).

This study has certain limitations. Although multicenter in design, the sample size remains relatively modest, and variability in MRI equipment and acquisition protocols across centers may have influenced the consistency of imaging measurements and model stability. Moreover, the analysis focused primarily on conventional MRI and DWI parameters, without integrating other functional modalities such as dynamic contrast-enhanced MRI, MR spectroscopy, or perfusion imaging. Future studies with larger cohorts, multimodal imaging, and longitudinal data are warranted to refine and validate the model’s clinical utility.

Conclusions

This study reaffirms the value of MRI-based imaging features in the histologic grading of STSs and introduces a novel, multi-parameter assessing model with strong potential for clinical translation. These findings lay the groundwork for future quantitative and standardized imaging approaches in tumor characterization and management.

Acknowledgments

None.

Footnote

Reporting Checklist: The authors have completed the STARD reporting checklist. Available at https://qims.amegroups.com/article/view/10.21037/qims-2025-1-2778/rc

Data Sharing Statement: Available at https://qims.amegroups.com/article/view/10.21037/qims-2025-1-2778/dss

Funding: This work was supported by Henan Province Science and Technology Key Project (No. 232102311052).

Conflicts of Interest: All authors have completed the ICMJE uniform disclosure form (available at https://qims.amegroups.com/article/view/10.21037/qims-2025-1-2778/coif). The authors have no conflicts of interest to declare.

Ethical Statement: The authors are accountable for all aspects of the work in ensuring that questions related to the accuracy or integrity of any part of the work are appropriately investigated and resolved. The study was conducted in accordance with the Declaration of Helsinki and its subsequent amendments. This study was approved by the Ethics Committee of Henan Cancer Hospital (approval No. 2023-KY-0063-001). The waiver of informed consent was granted because the study was retrospective in nature, and did not involve any additional interventions or procedures beyond routine clinical care.

Open Access Statement: This is an Open Access article distributed in accordance with the Creative Commons Attribution-NonCommercial-NoDerivs 4.0 International License (CC BY-NC-ND 4.0), which permits the non-commercial replication and distribution of the article with the strict proviso that no changes or edits are made and the original work is properly cited (including links to both the formal publication through the relevant DOI and the license). See: https://creativecommons.org/licenses/by-nc-nd/4.0/.

References

1. Mirza B, Weedon D. Atypical fibroxanthoma: a clinicopathological study of 89 cases. Australas J Dermatol 2005;46:235-8. [Crossref] [PubMed]
2. Iorizzo LJ 3rd, Brown MD. Atypical fibroxanthoma: a review of the literature. Dermatol Surg 2011;37:146-57. [Crossref] [PubMed]
3. Fretzin DF, Helwig EB. Atypical fibroxanthoma of the skin. A clinicopathologic study of 140 cases. Cancer 1973;31:1541-52.
4. Fletcher CD. The evolving classification of soft tissue tumours - an update based on the new 2013 WHO classification. Histopathology 2014;64:2-11. [Crossref] [PubMed]
5. Fletcher CD. Pleomorphic malignant fibrous histiocytoma: fact or fiction? A critical reappraisal based on 159 tumors diagnosed as pleomorphic sarcoma. Am J Surg Pathol 1992;16:213-28.
6. Lin X, Davion S, Bertsch EC, Omar I, Nayar R, Laskin WB. Federation Nationale des Centers de Lutte Contre le Cancer grading of soft tissue sarcomas on needle core biopsies using surrogate markers. Hum Pathol 2016;56:147-54. [Crossref] [PubMed]
7. Schneider N, Strauss DC, Smith MJ, Miah AB, Zaidi S, Benson C, van Houdt WJ, Jones RL, Hayes AJ, Fisher C, Thway K. The Adequacy of Core Biopsy in the Assessment of Smooth Muscle Neoplasms of Soft Tissues: Implications for Treatment and Prognosis. Am J Surg Pathol 2017;41:923-31. [Crossref] [PubMed]
8. Chhabra A, Ashikyan O, Slepicka C, Dettori N, Hwang H, Callan A, Sharma RR, Xi Y. Conventional MR and diffusion-weighted imaging of musculoskeletal soft tissue malignancy: correlation with histologic grading. Eur Radiol 2019;29:4485-94. [Crossref] [PubMed]
9. Crombé A, Marcellin PJ, Buy X, Stoeckle E, Brouste V, Italiano A, Le Loarer F, Kind M. Soft-Tissue Sarcomas: Assessment of MRI Features Correlating with Histologic Grade and Patient Outcome. Radiology 2019;291:710-21. [Crossref] [PubMed]
10. Xu W, Hao D, Hou F, Zhang D, Wang H. Soft Tissue Sarcoma: Preoperative MRI-Based Radiomics and Machine Learning May Be Accurate Predictors of Histopathologic Grade. AJR Am J Roentgenol 2020;215:963-9. [Crossref] [PubMed]
11. Yan R, Hao D, Li J, Liu J, Hou F, Chen H, Duan L, Huang C, Wang H, Yu T. Magnetic Resonance Imaging-Based Radiomics Nomogram for Prediction of the Histopathological Grade of Soft Tissue Sarcomas: A Two-Center Study. J Magn Reson Imaging 2021;53:1683-96. [Crossref] [PubMed]
12. Zhao F, Ahlawat S, Farahani SJ, Weber KL, Montgomery EA, Carrino JA, Fayad LM. Can MR imaging be used to predict tumor grade in soft-tissue sarcoma?. Radiology 2014;272:192-201. [Crossref] [PubMed]
13. Kim JY, Kim JJ, Hwangbo L, Kang T, Park H. Diffusion-weighted Imaging of Invasive Breast Cancer: Relationship to Distant Metastasis-free Survival. Radiology 2019;291:300-7. [Crossref] [PubMed]
14. Fernebro J, Wiklund M, Jonsson K, Bendahl PO, Rydholm A, Nilbert M, Engellau J. Focus on the tumour periphery in MRI evaluation of soft tissue sarcoma: infiltrative growth signifies poor prognosis. Sarcoma 2006;2006:21251. [Crossref] [PubMed]
15. Li X, Wang Q, Dou Y, Zhang Y, Tao J, Yang L, Wang S. Soft tissue sarcoma: can dynamic contrast-enhanced (DCE) MRI be used to predict the histological grade? Skeletal Radiol 2020;49:1829-38. [Crossref] [PubMed]
16. Soldatos T, Ahlawat S, Montgomery E, Chalian M, Jacobs MA, Fayad LM. Multiparametric Assessment of Treatment Response in High-Grade Soft-Tissue Sarcomas with Anatomic and Functional MR Imaging Sequences. Radiology 2016;278:831-40. [Crossref] [PubMed]
17. Li X, Hu Y, Xie Y, Huang B, Chen S, Tao H. Soft-tissue sarcoma: modified grading method improves the accuracy of preoperative MRI in predicting patient outcomes. Eur Radiol 2025;35:4571-86. [Crossref] [PubMed]
18. Hauwanga WN, McBenedict B, Goh KS, Yau RCC, Thomas A, Alphonse B, Ahmed YA, Yusuf WH, Devan JN, Alsiddig HA, Ahmed A, Lima Pessôa B. Magnetic Resonance Imaging Features for Distinguishing High-Grade From Low-Grade Soft Tissue Sarcoma: A Systematic Review and Meta-Analysis. Cureus 2024;16:e72784. [Crossref] [PubMed]
19. Fisher SM, Joodi R, Madhuranthakam AJ, Öz OK, Sharma R, Chhabra A. Current utilities of imaging in grading musculoskeletal soft tissue sarcomas. Eur J Radiol 2016;85:1336-44. [Crossref] [PubMed]
20. Schmitz F, Sedaghat S. Inferring malignancy grade of soft tissue sarcomas from magnetic resonance imaging features: A systematic review. Eur J Radiol 2024;177:111548. [Crossref] [PubMed]
21. Mali SA, Ibrahim A, Woodruff HC, Andrearczyk V, Müller H, Primakov S, Salahuddin Z, Chatterjee A, Lambin P. Making Radiomics More Reproducible across Scanner and Imaging Protocol Variations: A Review of Harmonization Methods. J Pers Med 2021;11:842. [Crossref] [PubMed]
22. Cheong EN, Park JE, Park SY, Jung SC, Kim HS. Achieving imaging and computational reproducibility on multiparametric MRI radiomics features in brain tumor diagnosis: phantom and clinical validation. Eur Radiol 2024;34:2008-23. [Crossref] [PubMed]
23. Rai R, Holloway LC, Brink C, Field M, Christiansen RL, Sun Y, Barton MB, Liney GP. Multicenter evaluation of MRI-based radiomic features: A phantom study. Med Phys 2020;47:3054-63. [Crossref] [PubMed]
24. Yu H, Tang B, Fu Y, Wei W, He Y, Dai G, Xiao Q. Quantifying the reproducibility and longitudinal repeatability of radiomics features in magnetic resonance Image-Guide accelerator Imaging: A phantom study. Eur J Radiol 2024;181:111735. [Crossref] [PubMed]