Plain radiographs showed phleboliths in 8 patients (34%; Figure 1). All lesions with phleboliths were located intramuscularly. On MRI, phleboliths were visualised in the form of nodular hypointense spots. The size of the focal lesions ranged from 1×1×2 cm to 6×11×14 cm. The lesions diffusely involved the extremities in four patients (Figure 2). Five patients (three of whom were children) had a lesion with a subcutaneous component; two of the children had both haemangioma and lymphangioma.

On T₁ weighted MR images, the lesions and neighbouring muscle tissue were isohyperintense in 13 patients, isointense in 7 patients, slightly hyperintense in 2 patients and hypointense in 1 patient (Figure 3). A hyperintense rim in the margin of the lesion was observed in three patients (Figure 3d).

On T₂ weighted MR images, all lesions had hyperintense signals. On T₂ weighted MR images, there were nodular hypointense areas in the lesions in 13 patients (Figure 4). There was lobulation and septation in 19 patients and septation alone in one patient. Two patients had a fluid–fluid level in the haemangioma (Figure 5). There was a neighbouring lesion with a haemosiderin ring in its membrane in one patient, suggestive of a haematoma in its chronic stage. The vessel supplying the lesion could be viewed in one patient.

Following injection of the contrast medium, 19 patients had heterogeneous enhancement patterns, while 3 patients had homogeneous enhancement patterns (Figure 6). One patient did not show contrast enhancement.

The lesion was in contact with the neighbouring bone cortex in 14 patients, 4 of whom had atrophy of the bone adjacent to the lesion. Of these four patients, one had a lesion extending from the elbow to the middle segment of the forearm involving both volar and dorsal compartments. Plain radiographs showed diffuse atrophy in one-third of the proximal radius, cortical thickening and trabecular coarsening in the ulna and soft tissue phleboliths. Another patient had an intramuscular lesion in the thigh with a subcutaneous component. The lesion was located in the anterior, medial and posterior compartments and the femur adjacent to the lesion had atrophy (Figure 7). Another patient had diffuse intramuscular lesions located in both the thigh and the lower leg. The lesions involved the anterior and posterior compartments. Both MRI and plain radiographs showed diffuse atrophy in the femur (Figure 8). The final patient had lesions unilaterally involving the entire lower extremity. The cutaneous, subcutaneous and intramuscular lesions involved all anatomical compartments of both the thigh and the lower leg. There was atrophy in the femur, expansion and sclerosis of the medullary bone distal to the fibula, a periosteal reaction of the lateral cortex of the tibia, thickening of the medial cortex of the tibia and sclerosis of the medullary bone (Figure 9). On MRI, none of the four patients had signal changes from the medullary bone. In the remaining 10 patients with lesions that were in contact with the bone, there were no reactive bone changes.

The diagnosis of soft-tissue haemangioma was based on excisional biopsy in 15 patients (65%) and on clinical signs and MRI findings in 8. Histological examination in 15 patients revealed that the tumour was cavernous haemangioma in 13 cases and arteriovenous haemangioma in 2. Of the four patients who underwent angiography, one had pre-operative embolisation and three had sclerotherapy. On pathological examination, two patients received a diagnosis of synovial haemangioma (Figure 10). Two children had haemangioma accompanied by lymphangioma. One child had had abdominal lymphangioma 1 year earlier. One patient with a lesion in the knee had a recurrence 3 years after surgery.

The lesions are located in the lower extremities in 45% of patients [9]. Patients with phleboliths (20–67%) have typical soft-tissue haemangioma [1,4]. Malignant transformation is rare [1,10]. Metastases of the lesions have not been reported [1]. No sex preponderance was stated in the study by Wild et al [9], although the lesions appeared more frequently in women [4] as reported in the present study. In 90% of patients, lesions are diagnosed in the first three decades of life [9]. In line with this observation, in 65% of our patients the lesions occurred in the first three decades of life.

Histologically, soft-tissue haemangioma can be classified into five types: capillary, cavernous, arteriovenous, venous and mixed haemangioma [1,3,4]. Capillary haemangioma is the most frequent haemangioma. This type of lesion is located in the cutaneous or subcutaneous tissues and is diagnosed in the first decade of life. Most instances of capillary haemangioma undergo involution spontaneously. Cavernous haemangiomas are large, deeply located and are diagnosed later in life; these lesions are frequently intramuscular, do not have spontaneous involution and require surgical treatment [4]. Arteriovenous haemangioma is composed of shunts and prevalence rates vary [4]; these lesions can be superficial or deeply located [4]. Venous haemangiomas are made up of clusters of large venous vessels with thick walls [11]. These lesions are typically located deep in the retroperitoneum, mesentery and the extremities [4]. Mixed haemangioma is, microscopically, a mixture of capillary and cavernous haemangioma [4,11].

Superficial haemangiomas are diagnosed easily because they cause discolorations of the skin and rarely require imaging techniques [1]. However, lesions deep in the tissues are not diagnosed easily because they grow slowly and do not cause discolorations of the skin. Imaging techniques are necessary to differentiate such haemangiomas from malignant lesions [1,2,7].

MRI is the standard imaging technique for diagnosing soft-tissue haemangioma [4]. On T₁ weighted images, compared with muscle tissue, intensities of the lesions are isointense or hyperintense with unclear margins [6^,7^,11-13]. In the present study, in keeping with the literature, T₁ weighted images showed haemangiomas to be isointense or hyperintense in 20 patients; however, in contrast with the literature, we found that two lesions were diffuse hyperintense and one was hypointense on T₁ weighted images. On T₂ weighted images, haemangiomas are typically hyperintense and have clear margins and lobulated contours. Although these signs are characteristic of haemangiomas, they are not pathognomonic [7]. As previously reported, the present study demonstrates hyperintensity on T₂ weighted images in all cases. In addition, all lesions had clear margins and most possessed lobulated contours. Marked hyperintensity of the lesions on T₂ weighted images is due to increased fluid content secondary to stagnant blood flow in large vessels [12,13]. Ehara et al [14] reported five cases with a fluid–fluid level in haemangioma. In the present study, two patients with intramuscular haemangioma located in the forearm had a fluid–fluid level.

Thin, linear, hypointense structures inside the lesions on T₂ weighted images result from fibrous septa between the vessels [12]. Moreover, occasional punctuate or reticular hypointense areas might be due to fibrous tissue, fast blood flow in vessels, calcification, ossification, haemosiderin, smooth muscle components or a thrombosis in vascular structures [2,11,13]. In such cases, plain radiographs and CT can be useful in differentiating calcification from ossification. In this study, 13 patients (56%) had nodular hypointense areas on T₂ weighted MRI images. Only eight of these patients had phleboliths on plain radiographs, a finding that led us to believe that not all nodular hypointensities on T₂ weighted images corresponded to phleboliths.

Intramuscular haemangiomas contain various amounts of fat, smooth muscle, myxoid stroma, thrombi and haemosiderin [11,12]. In some cases, haemangiomas contain so much fat that they can be mistaken for lipomas [4]. Lesions larger than 2 cm typically have different kinds of tissues and, therefore, have heterogeneous signals [4].

Deeply located, large haemangiomas sometimes cause changes in the neighbouring bone [1,3,8,10]. Ly et al [1] divided bone changes into three categories: periosteal, cortical and medullary. These authors classified the cortical changes into erosion, thickening, tunnelling and osteopenia; medullary changes were classified into osteosclerosis and trabecular coarsening. A correlation was indicated between bone changes and the distance between the lesion and the bone. Moreover, the authors reported medullary changes to be correlated not only with the proximity of the lesion to the bone, but also with the lesion's size [1]. Sung et al [8] classified bone changes into periosteal reaction, osteopenia, bone enlargement, cortical erosion, trabecular coarsening and a combination of all these. Some authors emphasise an enlargement in the neighbouring bone in cases of diffuse haemangiomas [8,15]. In the present study, four patients had atrophy in the neighbouring bone. The precise mechanism of reactive bone changes in soft-tissue haemangioma remains unknown. Several factors could contribute to the development of such bone changes, including an extrinsic pressure effect of the lesion, dilated vessels and hyperaemia secondary to vascularity of the tumour [16]. In our opinion, the mechanism for the development of bone atrophy might be related to the size, formation and duration of development of the mass prior to completion of bone growth. We believe that hypervascularity in the large and deeply seated soft-tissue haemangiomas might have caused disturbance of bone nutrition. In this study, in three of the four cases of bone atrophy, insufficiency of bone vascularisation could have resulted from the fact that the lesion was present in both subcutaneous and intramuscular compartments. Histopathological studies were undertaken of three of the four cases of bone atrophy; all three revealed cavernous haemangioma.

To our knowledge, in the English literature, there have been no reports of soft-tissue haemangioma associated with bone atrophy.

Although some researchers have noted that small haemangiomas rarely cause a periosteal reaction [7,10], Goto et al [3] claim this is not rare. These authors also proposed that a periosteal reaction could result from passive hyperaemia caused by a tumour, retraction or irritation of a tumour. Goto et al [3] claimed that a periosteal reaction was not stimulated by the size of the lesion, but that the distance between the lesion and the bone played an important role [3]. De Filippo et al [10] hypothesised that the periosteal reaction results from increased local vascularity caused by the tumour. There have been conflicting comments on the relationship between bone changes and pain. Ly et al [1] found no relation between pain and bone changes, whereas Goto et al [3] reported that patients with haemangiomas associated with periosteal reaction more frequently had pain. None of the four patients with bone changes in our study complained of pain.

Deeply located haemangiomas are mostly intramuscular, but they can also be synovial [13]. Intramuscular haemangiomas are frequently located in the trunk and lower extremities; these lesions rarely appear in the upper extremities [10]. Synovial haemangiomas are rare and almost always involve the knee joint; patients with synovial haemangiomas complain about pain, swelling and effusion; these lesions most frequently appear in the suprapatellar region [13]. Rare vascular tumours such as haemangioendothelioma and angiosarcoma, arteriovenous malformation, lymphangioma, lipoma, liposarcoma and other soft-tissue sarcomas might mimic haemangiomas [10,12,13].

Biopsy of soft-tissue haemangiomas can cause bleeding. In the present study, one patient with intramuscular haemangioma located in the forearm developed a spontaneous haematoma.

As for treating haemangiomas, symptomatic cases are treated with surgical resection or a laser [4,9,10]. Haemangioma has been reported to recur in 18% of patients who undergo surgery [5]. Mixed haemangioma tends to recur most frequently, followed by capillary and cavernous haemangiomas. In our study, haemangioma recurred in one patient 3 years after surgery. When excision is not appropriate, radiotherapy or embolisation can be useful [9]. We treated 15 patients with surgery, 3 with sclerotherapy under angiography and 1 with embolisation.