Diagnostic Performance of High-Frequency Ultrasound and Ultra-High-Frequency Ultrasound in Distinguishing Dermatofibrosarcoma Protuberans from Dermatofibroma: A 15-year Period Retrospective Analysis

Introduction

Dermatofibroma (DF) and dermatofibrosarcoma protuberans (DFSP) are two distinct cutaneous lesions that often present diagnostic challenges due to their overlapping clinical and imaging features.¹ DF, a benign fibrous histiocytoma, is one of the most common superficial skin lesions, typically presenting as a small, firm nodule.² In contrast, DFSP is a rare, low- to intermediate-grade sarcoma characterized by its locally aggressive behavior and high recurrence rate if not adequately excised.³ Accurate differentiation between these two entities is crucial for appropriate clinical management, as DFSP requires wide surgical excision to prevent recurrence, while DF can often be managed conservatively.^4,5

Ultrasound imaging, particularly high-frequency ultrasound (HFUS) and ultra-high-frequency ultrasound (UHFUS), has emerged as a valuable tool in the evaluation of superficial soft tissue lesions.⁶ HFUS provides excellent visualization of the dermis and subcutaneous tissue, while UHFUS, operating at even higher frequencies (typically 20–100 MHz), further improves spatial resolution, enabling more detailed characterization of cutaneous structures.⁷ This high-resolution imaging capability allows for the precise assessment of lesion depth, margins, and internal echogenicity, which are critical in distinguishing between benign and malignant cutaneous tumors.⁸

Previous studies have highlighted the utility of ultrasound in characterizing DF and DFSP. DF typically appears on ultrasound as an ill-defined, hypoechoic lesion in the dermis or upper hypodermis, often with spiculated margins and changes in adjacent soft tissue echogenicity.^9,10 In contrast, DFSP often presents as a hypoechoic mass with irregular borders, tentacle-like projections, and hypervascularity, reflecting its infiltrative growth pattern.^11,12 However, the diagnostic accuracy of ultrasound in differentiating these two entities, particularly using HFUS and UHFUS, has not been extensively studied over extended periods.^13,14 Large-scale, long-term retrospective studies specifically evaluating the diagnostic performance of HFUS and UHFUS in differentiating DF from DFSP are scarce.

This retrospective analysis aims to evaluate the utility of HFUS and UHFUS in distinguishing between DF and DFSP over 15 years. By examining a large cohort of patients, we seek to identify specific sonographic features that can aid in the accurate diagnosis and management of these lesions, thereby facilitating appropriate surgical planning and potentially reducing recurrence rates.

Materials and Methods

Study Design and Population

This retrospective study was conducted using data from the Pathology Report System of Dongguan People’s Hospital from November 2009 to July 2025. The initial cohort comprised 899 surgically treated and pathologically confirmed patients, including 691 diagnosed with DF and 208 with DFSP. The final analytical cohort was derived from this group and consisted of 334 patients, specifically 207 DF cases and 127 DFSP cases, for whom HFUS or UHFUS images were available in the Picture Archiving and Communication System (Figure 1).

Figure 1 Patient selection flowchart for the retrospective study.

The inclusion criteria comprised: (1) pathologically confirmed diagnosis of DF or DFSP; (2) availability of complete preoperative HFUS or UHFUS imaging data; (3) no previous treatment or intervention before ultrasound examination. Exclusion criteria were: (1) incomplete clinical or imaging records; (2) history of surgical excision or other treatments prior to ultrasound evaluation.

Acquisition of Ultrasound Images

Before each examination, patients were instructed to expose the skin lesions in the appropriate position. To enhance image quality, a generous amount of ultrasound gel was applied between the transducer and the lesion surface. All lesions were first scanned using HFUS for initial lesion localization and comprehensive assessment, providing greater penetration depth for evaluating deeper tissue involvement. Subsequently, UHFUS was primarily employed to depict superficial anatomical structures and lesion details with high resolution. Grayscale ultrasound images were acquired by scanning all lesions, followed by a color Doppler flow imaging to visualize the internal blood flow. The HFUS and UHFUS settings, including depth, gain, focus, and frequency, were carefully adjusted to ensure a clear depiction of the lesions. The soft tissues were meticulously scanned for any lesions, and their locations were precisely marked on the ultrasound image. Representative images were securely saved in the informatics database of the Picture Archiving and Communications System for further analysis and documentation.

Interpretation of Ultrasound Images

On grayscale ultrasound imaging, various characteristics of the lesions were evaluated, including size, shape (regular, irregular), margin (well-defined, ill-defined), internal echogenicity (hypoechoic, isoechoic, hyperechoic), homogeneity (homogeneous, heterogeneous), and posterior acoustic features (no change, attenuation, enhancement). Based on the aforementioned sonographic features, the lesions were further categorized into specific ultrasound patterns: compressive pattern, serrated pattern, tentacle-like border pattern, internal hyperechoic area, peripheral hyperechoic rim, mixed echogenicity pattern, all of which were illustrated in Figure 2.

Figure 2 Ultrasound patterns of dermatofibrosarcoma protuberans and dermatofibroma. Schematic illustrations (left) and corresponding grayscale ultrasound images (right) are provided for each pattern. (a and b) Compressive pattern: A well-defined, regularly shaped lesion exhibiting homogeneous echotexture. (c and d) Serrated pattern: An ill-defined lesion with serrated margins, showing homogeneous echotexture. (e and f) Tentacle-like border pattern: A lesion with tentacle-like extensions invading the subcutaneous tissue, displaying either homogeneous or heterogeneous echotexture. (g and h) Internal hyperechoic area: A lesion containing internal hyperechoic areas, which may appear regular or irregular in shape. (i and j) Peripheral hyperechoic rim: An irregular lesion surrounded by a hyperechoic rim. (k and l) Mixed echogenicity pattern: A lesion demonstrating multiple characteristic features, including tentacle-like extensions, internal hyperechoic areas, and a peripheral hyperechoic rim.

Color Doppler flow imaging was performed by superimposing color pixels onto grayscale ultrasound images to evaluate the presence and characteristics of blood flow in real-time. The diagnostic criteria for color doppler flow imaging were established as follows: Adler grade 0, indicating absence of blood flow and no vascularization; Adler grade 1, representing minimal vascularization with one or two pixels containing flow; Adler grade 2, demonstrating moderate vascularization with one main or several small blood vessels; and Adler grade 3, indicating rich vascularization with more than two main or four small blood vessels. The spatial distribution of the largest diameter blood flow was categorized into specific patterns (absent, arborizing, peripheral, random).

Statistical Analysis

Statistical analyses were performed using SPSS software (version 26.0, IBM Corp, Armonk, NY, USA). Continuous variables, such as age and lesion size, were presented as mean ± standard deviation and compared using independent samples t-tests. Categorical variables, including gender, lesion location, shape, margin, internal echogenicity, homogeneity, posterior acoustic features, ultrasound patterns, blood flow patterns, and blood flow grading, were expressed as frequencies and percentages. Inter-rater reliability for the Ultrasound pattern was evaluated using Cohen’s kappa. Group differences for categorical variables were assessed using the chi-square (χ²) test or Fisher’s exact test, as appropriate.

Univariate logistic regression analysis was conducted to identify potential predictors of DFSP and DF. We tested for collinearity using the variance inflation factor (VIF) and defined problematic collinearity as a VIF > 5. Variables with P-values < 0.05 in the univariate analysis were included in the multivariate logistic regression model to determine independent predictors of DFSP. Odds ratios (ORs) and 95% confidence intervals (CIs) were calculated to quantify the strength of associations. A two-tailed P-value < 0.05 was considered statistically significant.

Results

Patient Demographics and Clinical Characteristics

A total of 334 patients were included in the study, comprising 207 cases of DF and 127 cases of DFSP. The mean age of DF patients was significantly younger than DFSP patients (34.00 ± 12.270 years vs 40.99 ± 13.95 years; t = 4.794, P < 0.001). Gender distribution showed no significant difference between the two groups (χ² = 3.351, P = 0.067), with 96 males and 111 females in the DF group, and 72 males and 55 females in the DFSP group.

The anatomical distribution of lesions differed significantly between DF and DFSP (χ² = 43.309, P < 0.001). DF lesions were more commonly located on the limbs (head: 12 cases; girdle: 34 cases; limb: 101 cases), whereas DFSP lesions predominated in the trunk (77 cases) (Table 1).

Table 1 Patient Demographics and Clinical Characteristics

Ultra-High-Frequency and High-Frequency Sonographic Features

The mean lesion size was significantly larger in DFSP compared to DF (43.02 ± 39.47 mm vs 10.34 ± 7.07 mm; t = 9.240, P < 0.001). Morphologically, DF lesions were more often regular and well-defined, whereas DFSP lesions tended to be irregular (78 cases vs 58 cases; χ² = 36.371, P < 0.001) and ill-defined (63 cases vs 77cases; χ² = 4.977, P = 0.026). Internal echogenicity was predominantly hypoechoic in both groups, but DFSP showed higher heterogeneity (117 cases vs 79 cases; χ² = 94.527, P < 0.001). Posterior acoustic features also differed, with DFSP more frequently exhibiting enhancement (67 cases vs 61 cases; χ² = 20.038, P < 0.001). Distinct ultrasound patterns differentiated DF and DFSP lesions (χ² = 114.318, P < 0.001). DF lesions were predominantly either compressive (n=136) or serrated (n=38), whereas DFSP lesions most commonly exhibited tentacle-like borders (n=22), internal hyperechoic areas (n=30), a peripheral hyperechoic rim (n=15), or mixed echogenicity (n=24). Interobserver agreement for the ultrasound pattern classification was substantial interobserver agreement, with a Kappa value of 0.893 (P < 0.001).

Blood flow analysis demonstrated significant differences in vascular patterns. DF lesions were often avascular (162 cases), while DFSP lesions displayed random (67 cases), peripheral (21 cases), and arborizing (13 cases) distributions (χ² = 109.577, P < 0.001). Blood flow grading further distinguished DFSP, with higher Adler grades compared to DF (101 cases vs 45 cases; χ² = 107.999, P < 0.001) (Table 2).

Table 2 Ultra-High-Frequency and High-Frequency Sonographic Features

Multivariate Analysis of Patient Demographics, Clinical Characteristics, and Sonographic Features

The multicollinearity assessment showed that the variance inflation factor (VIF) values for all included variables were below 5. This confirms the absence of substantial multicollinearity among the predictors and supports their appropriateness for inclusion in the multivariate logistic regression model (Table 3). Multivariate logistic regression analysis revealed that tumor location, size, and ultrasound pattern were significant predictors of DF and DFSP. Specifically, girdle and limb (OR = 0.078, 0.079; P = 0.003, < 0.001) showed strong associations with DF, while size, tentacle-like border pattern, internal hyperechoic area, peripheral hyperechoic rim, and mixed echogenicity pattern (OR = 1.174, 4.391, 9.155, 9.328, 50.644; P < 0.001, 0.046, 0.002, 0.003, < 0.001) showed strong associations with DFSP. However, blood flow pattern did not demonstrate independent significance in the multivariate model (Table 4). The sensitivity, specificity, Youden’s index, and area under the curve of the multivariate model were 0.858, 0.971, 0.829, and 0.970, respectively (Figure 3).

Table 3 Multicollinearity Check of Patient Demographics, Clinical Characteristics, and Ultrasound Image Features

Table 4 Univariate and Multivariate Analysis of Patient Demographics, Clinical Characteristics, and Ultrasound Image Features

Figure 3 ROC curve of the multivariate model.

Discussion

Demographic and Clinical Distinctions

This study provides a comprehensive evaluation of HFUS and UHFUS in DF from DFSP, two cutaneous lesions that frequently present diagnostic challenges. Clinically, DF typically appears as small, firm, slow-growing nodules predominantly found on the extremities of young to middle-aged women.^15,16 In contrast, DFSP manifests as larger, progressively enlarging plaques or nodules primarily occurring in adults aged 30–50 years, with no gender predilection.^17,18 Anatomically, DF predominantly affects the head, girdle, limb, while DFSP favors the trunk. These patterns provide valuable diagnostic clues: extremity lesions with typical features likely represent DF, whereas truncal lesions should raise suspicion for DFSP.^19,20 These findings may reflect DF’s association with minor trauma (common on extremities) versus DFSP’s potential embryological origin (explaining truncal predominance). These clinical differences, while helpful, often require imaging confirmation for accurate diagnosis.

Sonographic Differentiation Features

Our imaging findings underscore sonographic features of DFSP and DF as visualized by HFUS and UHFUS. DF typically manifests as hypoechoic nodules with poorly defined margins, regular contours, and spiculated borders, often accompanied by altered adjacent tissue echogenicity, reflecting its benign fibrous proliferation and chronic inflammatory nature. These features correlate with its indolent clinical course, rarely requiring extensive intervention.²¹ In contrast, DFSP exhibits more aggressive features, including an irregular tentacle-like border and marked hypervascularity, reflecting its infiltrative growth pattern.²² The tumor frequently exhibits heterogeneous echotexture, posterior acoustic changes, and prominent vascularity with higher blood flow grades, all of which align with its malignant potential and propensity for local recurrence if not adequately resected. Notably, certain sonographic patterns, such as internal hyperechoic foci and mixed echogenicity, are more frequently associated with DFSP, while compressive patterns and homogeneous echotexture are more typical of DF.^23–25 The findings emphasize the utility of HFUS and UHFUS in characterizing these tumors, particularly in assessing morphological details, vascular patterns, and depth of invasion, which are critical for clinical management. Compared to HFUS, UHFUS allowed visualization of subtle architectural details that HFUS or earlier studies could not reliably depict. Specifically, UHFUS enhanced the delineation of fine internal reflectivity within lesions, improved border definition—particularly in small (<10 mm) or superficially located tumors—and more clearly revealed subtle tentacle-like projections infiltrating the subcutaneous septum. These capabilities are crucial for distinguishing early or small DFSP from benign mimics like DF, where margin assessment is critical.

Multivariate Predictive Modeling

Our multivariate analysis identified key predictors of DF and DFSP: location, size, and ultrasound patterns. This analytical strength differentiates our work from numerous prior studies that primarily reported univariate associations. For example, while previous studies noted the presence of features like tentacle-like projections or hypervascularity, our model quantified their independent predictive power while simultaneously by integrating other variables like tumor size and location. These findings align with established clinical risk factors and should prompt suspicion for DF or DFSP when present. Typical DF cases with characteristic clinical and sonographic features, such as girdle, limb, and small size, may be managed conservatively, while lesions showing DFSP characteristic clinical and sonographic features, such as large size, tentacle-like border pattern, internal hyperechoic area, peripheral hyperechoic rim, and mixed echogenicity pattern, warrant biopsy and wide excision with 2–3 cm margins due to their aggressive potential.²⁶ For different anatomical sites, selecting the appropriate pedicled or free flap for reconstruction after wide excision ensures good functional and aesthetic outcomes with no local recurrence.^27–30

Both HFUS and UHFUS aid preoperative planning, which achieve 81.8% sensitivity for detecting muscle invasion, potentially reducing incomplete excisions and recurrence.³¹ From a clinical management perspective, the multivariate model provides a practical tool for risk stratification. The strong association of girdle and limb location with DF suggests that ultrasound identification of a small lesion on the girdle or limb may allow for watchful waiting in clinically typical cases. In contrast, the identification of a large, lesion on the trunk with a tentacle-like border pattern, internal hyperechoic area, peripheral hyperechoic rim or mixed echogenicity pattern should be considered highly suspicious for DFSP, mandating definitive histological diagnosis and surgical consultation. This predictive model can help clinicians prioritize higher-risk lesions for expedited intervention, potentially improving outcomes by reducing time to treatment for aggressive tumors.

Our systematic classification of DF and DFSP using HFUS and UHFUS revealed strong correlations with histopathology and clinical behavior. The localized pattern (65.2% of DF cases) appears as a well-defined hypoechoic nodule, consistent with circumscribed growth. The serrated pattern (18.4% of DF cases) shows heterogeneous echogenicity with serrated structures, requiring differentiation from early DFSP.³² The tentacle-like border pattern, more frequent in DFSP (17.3% of DFSP cases), is caused by tumor cells spreading along the fat septum, indicating infiltrative growth and necessitating wide excision.^33,34 Peripheral hyperechoic rim (11.8% of DFSP cases) likely indicates involvement of surrounding adipose tissue by the lesion, while the internal hyperechoic pattern (23.6% of DFSP cases) represents the invasion and encasement of adipose tissue by the lesion, both of which likely reflect collagen deposition or necrosis, enhancing diagnostic specificity.^10,35 The mixed echogenicity pattern (18.9% of DFSP cases) underscores tumor heterogeneity, likely stemming from the intricate interweaving of tumor cell bundles and collagen fibers.

Limitations and Future Directions

This study has several limitations that should be considered when interpreting the results. As a single-center retrospective analysis spanning 15 years, the extended timeframe may have introduced variability in image acquisition and interpretation. It is also important to note that a small subset of patients did not undergo ultra-high frequency ultrasound examination. Although the initial cohort was large, the final analytical sample was substantially reduced, partly due to limited clinician familiarity with cutaneous ultrasound which led to the underutilization of such examinations prior to lesion excision, raising the possibility of selection bias. Furthermore, it is important to note that the initial univariate screening involved a substantial number of comparisons without formal statistical correction, which may increase the risk of Type I error (false positives). These univariate results were used solely for the purpose of informing variable selection for the subsequent multivariate model and should be interpreted with this caveat in mind.

Future research should aim to address these limitations. Prospective, multi center studies using standardized ultrasound protocols and equipment are needed to validate our findings and enhance their generalizability. Efforts to increase clinician awareness and training in cutaneous ultrasound could help mitigate selection bias in future patient recruitment. Furthermore, the integration of advanced multimodal ultrasound techniques, such as elastography and contrast enhanced ultrasound, which the existing literature suggests hold significant potential for improving diagnostic accuracy, should be prioritized.^34,36,37 Finally, incorporating artificial intelligence and explainable artificial intelligence models could help overcome challenges related to variable interpretation and multicollinearity by standardizing image analysis and identifying the most salient diagnostic features.³⁸ By combining refined ultrasound examinations with these emerging technologies, future work can advance the development of robust, personalized diagnostic strategies to improve patient outcomes.

Conclusion

This 15-year retrospective analysis demonstrates that ultrasound, particularly at higher frequencies, is a powerful, non-invasive tool for characterizing dermatofibrosarcoma protuberans (DFSP) and dermatofibroma (DF). DFSP lesions are typically larger, located on the trunk, and exhibit distinct sonographic features, including heterogeneity, tentacle-like borders, peripheral hyperechoic rims, mixed echogenicity, and marked hypervascularity. In contrast, DF is more common on the extremities, is smaller, and often appears homogeneous with compressive or serrated patterns. Multivariate analysis confirmed that tumor location, larger size, and specific ultrasound patterns are independent predictors. These sonographic findings provide critical preoperative diagnostic information, guiding appropriate clinical management with conservative approaches for DF and wide surgical excision for DFSP. Future multi-center studies are warranted to validate these criteria, and the integration of artificial intelligence could further enhance diagnostic standardization and accuracy.