Extraskeletal osteosarcoma (EOS) is a malignant mesenchymal tumor that produces osteoid or bone matrix but arises entirely within soft tissue or visceral organs rather than from any bony structure. It shares the same histomorphological characteristics as conventional intraosseous osteosarcoma, yet its biological behavior, clinical presentation, and prognosis differ considerably from tumors that originate in bone. EOS accounts for roughly 1% of all soft tissue sarcomas and approximately 4% of all osteosarcomas, placing it firmly in the ultra-rare category where systematic imaging data are scarce and individual clinician experience is limited.
The first formal description of EOS is attributed to Wilson in 1941, yet more than 80 years later its etiology remains incompletely understood. The most widely accepted hypothesis invokes chemical-biological transformation: fibroblasts within muscle or connective tissue are stimulated by internal or external factors to undergo metaplastic differentiation into osteoblast-like or chondroblast-like cells, which then undergo malignant transformation. Prior irradiation, trauma, and underlying conditions such as Paget's disease have all been implicated as risk factors in individual case reports, though no single causal mechanism dominates the literature.
Because EOS is not anchored to bone, it can appear in almost any anatomical location, ranging from the deep soft tissues of the limbs to the abdominal cavity, uterus, lung, liver, and bladder. This anatomical versatility means that any imaging department is likely to encounter the tumor in an unexpected site, making pattern recognition all the more important. The nonspecific clinical presentation, which typically consists of a slowly enlarging painless mass without constitutional symptoms, further delays diagnosis. Accurate, timely identification through CT and MRI is therefore a critical step in the clinical pathway, and this paper provides one of the more granular radiological-pathological correlation datasets yet published for this disease.
This study is a retrospective, multi-institutional case series spanning January 2008 to March 2023, drawing on pathologically confirmed EOS cases from two hospitals in Ganzhou, China. The ethics committee of Ganzhou People's Hospital (The Affiliated Ganzhou Hospital of Nanchang University) approved the study and waived the requirement for informed consent given the retrospective design. Eleven patients with histologically verified EOS were identified over the 15-year period, reflecting the extreme rarity of the diagnosis. All clinical manifestations, imaging results, and pathology reports were extracted for each case.
CT acquisition: At one institution, CT was performed on a GE Revolution scanner using 5 mm slice thickness and spacing, 120 kV tube voltage, and 382 mA tube current; multiplanar reconstruction (MPR) images were generated at the workstation. The contrast agent was nonionic iohexol (350 mg I/mL) at 1.2 mL/kg and a flow rate of 3.0 mL/s. At the second institution, a Siemens SOMATOM Definition dual-source scanner was used under analogous parameters (120 kV, 526 mA, 5 mm slices). Both sites generated MPR reconstructions to enable coronal and sagittal analysis of mass morphology.
MRI acquisition: One center used a Siemens Verio 3.0T scanner with fast spin echo sequences. T1-weighted images (T1WI) were acquired at TR 500 ms / TE 15 ms, and T2-weighted images (T2WI) at TR 3000 ms / TE 90 ms. Fat suppression T2WI and diffusion-weighted imaging were obtained where clinically indicated. The second center used a GE Signal Excite 3.0T superconducting system with spin-echo sequences and coil selection tailored to the anatomical site of interest. Gadolinium-based contrast enhancement was applied across both sites.
Image interpretation: Two senior radiologists independently reviewed and recorded the location, shape, size, density or signal characteristics, enhancement pattern, calcification or ossification morphology, and evidence of necrosis, cystic degeneration, or hemorrhage. Their findings were then correlated with final histopathological diagnoses. No formal inter-rater reliability metric (such as Cohen's kappa) is reported in the paper.
In this cohort of 11 patients, 6 were male (55%) and 5 were female (45%), with ages ranging from 23 to 76 years and a mean age of 47.1 years. This demographic profile is consistent with the wider literature, which identifies EOS as predominantly a disease of middle-aged to older adults, in contrast to conventional intraosseous osteosarcoma, which peaks in adolescence. The slightly male-predominant distribution also aligns with prior case series, though the small sample size precludes firm conclusions about sex-based risk differences.
Anatomical distribution: Limb and limb-girdle sites dominated in this series. Two patients had EOS in the thigh (one in the space between posterior muscles of the lower thigh, one between lateral muscles), two had EOS in the buttocks (left and right), and one had EOS in each of the following locations: the lesser omental bursa (minor omentum region), the uterus, the abdominal cavity, the ileocecal region, the chest wall, and the knee. This spread illustrates the genuinely panoptic distribution of EOS and the need for imaging teams to maintain the diagnosis on their radar regardless of anatomical site.
Laboratory findings: Serum alkaline phosphatase (ALP) was elevated before treatment in four patients, with the highest recorded value reaching 615 U/L against a normal reference range of 45-125 U/L. The remaining four patients had ALP within normal limits prior to treatment; ALP was not measured in the other three cases. Post-treatment ALP levels declined in two patients but remained above the normal range (227 U/L and 158 U/L), while one patient returned to normal after treatment. Lactate dehydrogenase (LDH), another recognized prognostic marker for osteosarcoma, was not systematically reported in this cohort. The clinical onset was typically insidious, with a painless, slowly enlarging mass as the dominant symptom and no reliable constitutional features to prompt earlier evaluation.
CT was performed in 8 of the 11 patients and constitutes the primary cross-sectional modality described in this paper. The hallmark CT feature of EOS is calcification or ossification within the soft tissue mass. Seven of the 11 patients demonstrated clear calcifications or ossification on imaging, presenting in variable morphologies that ranged from coarse granules to patchy deposits to large, dense, lumpy or clump-like mineral foci. Two patients showed massive mature bone tumor formation resembling densely ossified cortical bone. Calcification was distributed from the central portions of the mass to the periphery in some patients, while others showed predominantly peripheral distribution, a pattern that differs from the characteristic peripheral-to-central zoning of myositis ossificans and can therefore contribute to differential diagnosis.
Density and enhancement characteristics: On plain CT, EOS masses typically appeared as heterogeneous, iso- to mixed-density soft tissue masses. Ten of the 11 patients showed varying degrees of necrosis and cystic degeneration within the tumor on imaging, and 2 of those 10 had concurrent hemorrhage within the mass. The necrotic-cystic components produced areas of low attenuation, while viable tumor parenchyma was isodense or slightly hyperdense relative to muscle. On contrast-enhanced CT, the viable parenchymal component showed gradual progressive enhancement, a pattern described as "obvious gradual enhancement" in the reported cases with enhancing tumors, while necrotic and cystic areas remained non-enhancing. This enhancement pattern, although not specific to EOS, helps delineate the viable tumor margin from necrotic core when planning biopsies or resections.
Abdominal and visceral EOS: The two visceral cases (uterine EOS and abdominal cavity EOS) demonstrated particularly dramatic CT findings, including large masses with extensive central cystic-necrotic regions and massive calcification or ossification occupying significant portions of the tumor volume. The uterine case also showed abdominal metastatic lesions with multiple calcifications at the time of presentation. The ileocecal EOS also exhibited calcification and cystic degeneration. These visceral cases highlight how radiologists interpreting abdominal CT must include EOS in their differential when confronted with a heterogeneous soft tissue mass containing mineralization, even when bone is nowhere near the site of origin.
MRI was performed in 6 of the 11 patients, providing superior soft tissue contrast and, in several cases, additional diagnostic information beyond what CT alone could offer. On T1-weighted sequences (T1WI), the parenchymal component of EOS masses was consistently close to muscle signal intensity, reflecting the moderately cellular, fibrous nature of viable sarcoma tissue. On T2-weighted sequences (T2WI), the masses showed mixed signal intensity, with the parenchymal component demonstrating slightly prolonged T2 signal (isointensity to mild hyperintensity relative to muscle) and the cystic-necrotic regions appearing as areas of markedly increased T2 signal. Fibrous septa, when present, were hypointense on both T1WI and T2WI, creating an internal network of low-signal bands within the tumor that can occasionally mimic the appearance of myxoid liposarcoma or other septated soft tissue tumors.
Capsule and border features: Most masses demonstrated relatively well-defined boundaries on MRI, and some showed complete or incomplete pseudocapsules. The capsule was understood to be composed of tumor cells, fibrous tissue, and inflammatory components generated by the tumor-host interaction, rather than a true anatomical capsule. Several cases showed evidence of peripheral invasion, including one case where the common peroneal nerve was encased by the mass. The presence of a pseudocapsule on MRI does not guarantee respectability or benign behavior and should not be used to exclude malignancy.
Fluid-fluid levels and hemorrhage: Two patients with hemorrhage within the mass showed the "fluid-fluid level" sign on MRI, in which layered blood products of different ages produce a horizontal interface between two signal zones within the same cystic space. This finding is also seen in aneurysmal bone cysts and telangiectatic osteosarcoma and, when identified in a soft tissue mass, should prompt consideration of EOS among the differential diagnoses. Mineralized components (calcification or ossification) appeared as areas of signal void or markedly hypointense foci on all MRI sequences, consistent with the lack of free protons in mineralized matrix.
Enhancement on MRI: Post-gadolinium T1WI consistently showed uneven, marked enhancement of the viable parenchymal component, reflecting the tumor's high vascularity within the non-necrotic portions. Cystic and necrotic zones did not enhance, producing a characteristic rim or nodular enhancement pattern in many cases. This heterogeneous enhancement pattern on MRI mirrors the CT enhancement behavior and reflects the fundamental tumor biology: viable sarcoma parenchyma surrounded by regions of outpaced vascular supply and consequent necrosis.
Histopathology remains the definitive standard for EOS diagnosis, and the authors provide detailed gross and microscopic descriptions for the 11 cases in their series. Grossly, the tumors appeared as lobulated masses with either well-defined or ill-defined borders; cut surfaces were gray-white to gray-red, with focal hemorrhage and necrosis visible to the naked eye. Several tumors showed a firm or gritty consistency at the cut surface corresponding to the mineralized areas seen on CT.
Microscopic architecture: Under the microscope, EOS demonstrates a wide range of highly atypical, pleomorphic cells distributed in a disorderly fashion throughout the tumor stroma. Three principal cell types can be identified: spindle-shaped cells resembling fibroblasts, cells embedded in osteoid matrix, and cells associated with cartilaginous differentiation. The ratio of these three components varies considerably from case to case, accounting in part for the heterogeneous imaging appearance. The diagnostic anchor is the presence of osteoid matrix, which forms a disorderly, fine branched lace-like structure around the tumor cells, distinct from the organized trabecular pattern of normal bone or the structured matrix of benign osteoblastic lesions.
Diagnostic immunohistochemistry: Although the paper does not detail a comprehensive IHC panel for all cases, it notes that the disorderly, fine branching lace-like osteoid structure is the morphological feature that distinguishes EOS from undifferentiated pleomorphic sarcoma (which may contain focal osteoid but without this architectural pattern) and from other ossifying soft tissue tumors. Cartilaginous tissue was seen in some tumor areas, indicating chondroblastic differentiation, which again mirrors the histological subtypes seen in intraosseous osteosarcoma. The combination of high nuclear pleomorphism, mitotic activity, osteoid production, and absence of bone of origin is necessary for the diagnosis.
The imaging features of EOS overlap substantially with several benign and malignant soft tissue conditions, and the authors devote considerable attention to helping radiologists navigate this differential. The most practically important distinctions are discussed below.
Myositis ossificans: This non-neoplastic condition results from ectopic bone and cartilage formation near muscle and bone, typically after trauma. Its critical distinguishing feature on imaging is the "zoning phenomenon": mature, well-organized lamellar bone appears at the periphery of the lesion, with an intermediate zone of osteoid, and a central core of immature non-ossifying fibroblastic cells. On CT, myositis ossificans therefore shows peripheral mineralization that is denser and better organized than the central or diffuse calcification of EOS. On MRI, the peripheral ossification creates a well-defined low-signal ring encasing a higher-signal center. This peripheral-to-central gradient is the opposite of what is typically described for EOS and is a useful discriminating feature in practice.
Extraskeletal chondrosarcoma: Most soft tissue chondromas and chondrosarcomas arise in the fingers and feet, though any site is possible. The key imaging signature is "ring-and-arc" cartilage matrix mineralization on CT, reflecting peripheral mineralization of rounded hyaline cartilage nodules, combined with a high water content signal on T2WI in the unmineralized cartilage portions. EOS can contain chondroblastic areas but does not produce the ring-and-arc pattern; its mineralization is typically central, irregular, or amorphous. Histologically, EOS contains osteoid in addition to any cartilaginous areas, while pure chondrosarcoma does not produce osteoid.
Synovial sarcoma: This tumor is more common in young and middle-aged adults and typically arises near large joints. Its imaging signature on T2WI is the "triple signal intensity" pattern: low signal (fibrous tissue), isointense signal (hemorrhage), and high signal (cystic component), relative to fat. Calcification is common in synovial sarcoma and can appear as dense, irregular foci similar to EOS. Key distinguishing features include the juxta-articular location, the younger age group, and the biphasic histological appearance with both epithelial and spindle cell components. Abdominal EOS also needs to be distinguished from gastrointestinal stromal tumors (GIST) and omental lesions, which generally do not produce osteoid matrix.
Undifferentiated pleomorphic sarcoma: This is perhaps the most challenging differential for EOS because both tumors are high-grade, pleomorphic, and can contain focal osteoid or bone. The distinction rests on adequate tissue sampling: EOS shows the characteristic disorderly fine-branched lace-like osteoid throughout the tumor, whereas undifferentiated sarcoma shows only focal or reactive osteogenesis without this architectural pattern. On imaging alone, the two tumors may be indistinguishable.
Treatment of EOS mirrors the approach used for intraosseous osteosarcoma in many centers. In this cohort, 8 of the 11 patients underwent wide (radical) surgical excision, 1 underwent marginal excision, and 2 underwent open excisional biopsy alone due to disease extent or patient condition. Nine patients received chemotherapy after surgical resection, and one also received radiotherapy in addition to chemotherapy. Among those whose surgical margin status was documented, 2 patients had positive margins (R1 resection) and 3 had negative margins (R0); margin status was not recorded for the remaining cases, a significant limitation the authors themselves acknowledge.
Outcomes: Four patients died during follow-up (follow-up duration ranging from 2 to 49 months), yielding an observed survival rate of approximately 63% in this cohort, though the follow-up duration was highly variable and several patients were lost to follow-up. Among the deaths: one patient died of respiratory failure 2 months post-operation, two patients who underwent biopsy-only died of disease progression, and one died of lung metastasis 19 months after surgery (having developed lung metastases 6 months post-operatively). Two patients already had distant metastases at the time of initial diagnosis, confirming the aggressive biology of EOS. Recurrence was documented in one patient at 2 months after surgery.
Prognostic biomarkers: The authors discuss the prognostic significance of serum alkaline phosphatase and lactate dehydrogenase. Meta-analytic data cited in the paper demonstrate that elevated ALP and LDH levels are significantly associated with reduced overall survival in osteosarcoma. In this series, 2 of the 4 patients with elevated ALP died (at 2 and 11 months of follow-up), and 1 survived with disease at 5 months. Of the 4 patients with normal ALP, 2 were alive without disease progression at follow-up. While these numbers are too small to draw statistical conclusions, they are consistent with the published literature on ALP as a prognostic marker. Published 5-year overall survival rates for EOS range from 25% to 77% across larger series.
Limitations and future directions: This study carries several important limitations that the authors explicitly flag. First, the cohort of 11 patients is very small, reflecting the genuine rarity of the disease but limiting statistical inference. Second, follow-up duration was inconsistent and several patients were lost to follow-up, preventing reliable survival estimation. Third, surgical margin data were missing for several patients, which is critically important for interpreting recurrence and survival outcomes. Fourth, the retrospective, multi-center design introduced variability in imaging protocols, surgical management, and follow-up procedures between institutions. Fifth, chemotherapy regimens varied across patients and did not follow a uniform protocol, making it impossible to assess treatment efficacy systematically. Larger prospective registries or international collaborative databases will be needed to develop imaging-based predictive models and refine surgical and systemic treatment recommendations for this rare disease.