The committee proposes a bone reporting and data system (Bone-RADS) with four possible diagnostic management recommendations (Table 1): Bone-RADS1, leave alone; Bone-RADS2, perform different imaging modality; Bone-RADS3, perform follow-up imaging; and Bone-RADS4, biopsy and/or oncologic referral. If a lesion is clearly benign, such as in the case of an enostoses or non-ossifying fibroma, then it can and should be left alone (Bone-RADS1). If the lesion has clearly concerning features (large soft tissue component, intralesional enhancement) or is at risk for pathologic fracture, then the lesion needs to undergo biopsy and/or oncologic evaluation (Bone-RADS4). For indeterminate lesions, two options exist for the radiologist. The first option is to obtain different imaging to provide additional diagnostic information (Bone-RADS2), and the second option is to repeat the same imaging after a certain time period (Bone-RADS3) to assess for increase in size or development of worrisome features, such as cortical involvement. The recommendation is for Bone-RADS3 lesions 3 to undergo follow-up at 6,6,12 months intervals for a total of 2 years. For any bone lesion, one of these four management options should be selected. A major goal of the panel when creating these algorithms was to be certain that lesions designated Bone-RADS1 are truly benign processes that require no additional workup, such as a non-ossifying fibroma, enostoses, or focus of red marrow. Thus, the algorithms are weighted towards identifying Bone-RADS1 lesions.
Table 1 BONE-RADS categories for incidental bone lesions Three management algorithms are provided, two for CT lesions (“lucent” and “sclerotic/mixed”) and one for MRI. Both CT and MRI have useful features that can be used to aid the radiologist in arriving at the most appropriate recommendation. The CT algorithms incorporate lesion density, in Hounsfield units (HU), since very dense lesions (greater than 885 HU) and lesions containing fat (less than − 30 HU) are often benign [18,19,20,21]. Please note that there is some variation in the literature about the density range of fat in the literature. The most commonly accepted range is − 120 to − 30 HU, although one study has − 10 HU at its upper limit. Also, there is likely variability between patients, among different locations of the body, even within the same patient, and also due to imaging technique on different scanners, and therefore comparison to surrounding normal fat is often helpful [22,23,24]. The authors recognize that assessing the lesion density can be challenging for some lesions. Thus, the flowcharts have been designed to have overlap such that in indeterminate cases, either chart will result in the same diagnosis/recommendation. For MRI, we incorporate chemical shift imaging to identify microscopic fat within the lesion (Fig. 5) and post-contrast imaging to distinguish cyst-like lesions from solid lesions. In each algorithm, we have also incorporated whether the patient has a history of known malignancy with propensity to bone metastases (i.e., kidney, prostate, breast, lung, thyroid, assuming that the patient has had appropriate screening for these malignancies, such as prostate-specific antigen or screening mammography) and/or the presence of pain attributable to the lesion, as these clinical features would make the lesion more suspicious for malignancy.
Many of these incidental lesions seen on MRI or CT will be performed to evaluate painful conditions such as trauma, muscle injuries, stress fractures, or internal joint derangement and are often unrelated to the working clinical diagnosis, and orthopedic consultation may be required to clarify the source of pain. For example, if an incidental bone lesion is found on a shoulder MRI in a patient with a rotator cuff tear, the patient’s pain may be due to the lesion but could also be due only to the rotator cuff tear. Many entities can have variable imaging appearance and so could be included in more than one or even all three charts. Lastly, we want to stress that radiographs can be helpful in lesion assessment especially for incidental lesions discovered on MRI and should be taken into account if available.
Solitary lucent lesion on CT
The initial evaluation of a solitary lucent lesion on CT should begin with an assessment for concerning clinical (pain attributable to the lesion) or imaging (cortical involvement, soft tissue extension, pathologic fracture, or aggressive periosteal reaction) features [25]. The presence of any of these concerning features (Fig. 6) results in a Bone-RADS4 designation, prompting biopsy and/or oncologic referral for further workup. The term cortical involvement would include lesions with cortical tunneling, endosteal scalloping, bone remodeling/expansion, and cortical thickening. Lesions with these concerning features could include bone metastases, multiple myeloma, Langerhans cell histiocytosis, and osteomyelitis. While some features, particularly cortical involvement, may seem equivocal, referral for further evaluation with an orthopedic oncologist is often a safe and reasonable choice, as benign lesions, such as simple bone cysts, may also require treatment or prophylactic fixation to prevent pathological fracture.
For lesions without concerning features, the next step is to determine if the patient has a known malignancy with propensity to metastasize to the bone. If present, further evaluation with FDG-PET, bone scan, or MRI for further characterization (Bone-RADS2) or follow-up imaging (Bone-RADS3) to evaluate for interval change is recommended to minimize the risk of missing a new metastasis. It is important to restate that the algorithms only apply for solitary lesions and not multiple lesions.
If there are no concerning features or patient history of malignancy, the next step is to evaluate for significant fat in the lesion as several benign entities contain fat. Confirmation of the presence of fat can be directly visualized as macroscopic fat or determined by using a region of interest (ROI) placed in the lesion that measures with fat attenuation (mean density less than − 10 HU). If internal fat in the lesion is definitively confirmed, then the lesion is designated a Bone-RADS1, and no additional workup is needed. Intraosseous lipoma, hemangioma, and red marrow would be representative lesions.
Next, it is helpful to identify any benign lesions that can be recognized on CT with high confidence, specifically those with well-established characteristic features, such as fibrous dysplasia, non-ossifying fibroma, enchondroma, subchondral cyst, and hemangioma (Fig. 7). If a lesion is recognized as entirely consistent with any of these five benign entities, it is a Bone-RADS1 and requires no further workup. It should be reemphasized that if the lesion has concerning features such as endosteal scalloping in an enchondroma located in a long bone, then that lesion would have been designated a Bone-RADS4 at the beginning of the algorithm.
Any lesion that is not recognized as belonging in the Bone-RADS1 management category would require further workup. The urgency of the workup is based on the imaging features. For cortically based lesions, radiologists should assess for features suggesting an osteoid osteoma, osteoblastoma, or cortical metastasis. Medullary lucent appearing lesions encompass a variety of benign and malignant entities, including aneurysmal bone cyst, unicameral bone cyst, giant cell tumor, chondroblastoma, clear cell chondrosarcoma, myeloma, or metastasis. All of these lesions should be classified as Bone-RADS4 and referred to oncology and/or biopsy to prompt definitive management. If the lesion lacks clear features, it should either undergo additional workup with FDG-PET, bone scan, or MRI or be followed with additional imaging in 6 months.
Solitary sclerotic/mixed density lesion on CT
Similar to the approach of a lucent lesion, the initial evaluation of a sclerotic/mixed density lesion on CT should begin with an assessment for concerning clinical (pain attributable to the lesion) or imaging (cortical involvement, soft tissue extension, pathologic fracture, or aggressive periosteal reaction) features. The presence of any of these concerning features results in a Bone-RADS4 designation and would include metastases, myeloma, chondrosarcoma, or infection. If the lesion lacks concerning features, then the next step is to determine if the patient has a known malignancy with propensity to metastasize to the bone. If there is a malignancy history, then further evaluation of the solitary lesion with FDG-PET, bone scan, or MRI for further characterization (Bone-RADS2) or follow-up imaging (Bone-RADS3) is warranted as with the lucent lesion CT algorithm.
For lesions without concerning features and in a patient without a known malignancy with propensity to metastasize to the bone, the lesion should be assessed for whether it has any features that are highly characteristic of a benign sclerotic bone lesion, such as non-ossifying fibroma, enostosis, osteoma, or sclerosing bone dysplasia. If the classic features for one of these benign entities are present, the lesion can be designated Bone-RADS1 and left alone. Regarding enostosis, studies have shown that if a purely sclerotic lesion demonstrates a mean HU > 885, this is sensitive and specific for enostosis [18, 20, 25, 26]. However, in the setting of an older patient, reliance on thresholds should be limited. Interestingly, a recent study from Hong et al. showed that expert readers and radiomics models outperform strict HU thresholds[27].
If the lesion is not purely sclerotic, the presence of internal fat should be assessed. If present, then the lesion is designated a Bone-RADS1 and can be left alone. Differential considerations when internal fat is present include intraosseous lipoma, hemangioma, Paget disease, osteonecrosis, fibrous dysplasia, or non-ossifying fibroma (NOF).
If no fat is present, the lesion should then be assessed for “ground glass” appearance of fibrous dysplasia. The term “ground glass” has become synonymous with fibrous dysplasia in skeletal lesions, but it probably found its origins in thoracic radiology, where it described a diffusely hazy and smooth appearance. The range of densities varies widely, from sclerotic to lucent [28]. If present, the lesion is designated Bone-RADS1, and no additional management is needed. Most fibrous dysplasia is isolated, but if there are endocrine abnormalities or other sites of skeletal deformity, consider bone scan or skeletal survey to look for other sites of disease.
If it does not have a “ground glass” appearance, the lesion should next be assessed for cartilage matrix which appears as punctate calcifications often as small arcs and rings. If cartilage matrix is present, one should determine whether the lesion has any imaging features suggesting aggressive growth, which include endosteal scalloping, expansile remodeling, cortical breakthrough, periosteal reaction, and/or soft tissue mass. For endosteal scalloping in a long bone, involvement of greater than two-thirds of the cortical thickness or extension of scalloping greater than two-thirds of the craniocaudal length of the lesion raise suspicion for malignancy [29]. An additional feature that may raise concern is incomplete mineralization because malignant cartilaginous lesions more often show areas that are not mineralized as compared to non-aggressive cartilaginous lesions (case 4, Fig. 14) [30, 31]. However, this should be considered in the context with the other features of the lesion and not as a stand-alone feature of aggression. Similarly, studies have shown overlap in the size range of enchondromas and chondrosarcomas of long bones [29, 32]. Therefore, size should not be used as a stand-alone discriminator for aggressiveness. Although rare, an axial skeleton location, such as the pelvis, ribs, and sternum, is also a concerning feature because enchondromas are very rare (< 1%) in the flat bones [33]. In fact, in the authors’ experience, axial lesions are generally referred to as low-grade chondrosarcomas during radiology-pathology correlation even when they lack aggressive imaging features. However, because low-grade lesions without aggressive growth features carry a good prognosis, these can be treated with close clinical and imaging follow-up or with intralesional curettage [34]. Lastly, an epiphyseal location raises concern for malignancy due to the predilection of clear cell chondrosarcoma in this location. If any of these concerning features are present, the cartilage lesion is designated a Bone-RADS4, and representative lesions include atypical cartilaginous tumor/grade I chondrosarcoma, chondrosarcoma, and chondroblastoma. If there are no concerning features, the cartilage forming lesion is presumed to be an enchondroma and can be deemed Bone-RADS1 and left alone. Due to the low rate of chondrosarcoma in incidental painless cartilaginous lesions of the long bones, routine follow-up is considered unnecessary and can be instead reserved for when new symptoms arise[32, 35, 36].
Assessment of lesion density
One of the advantages of CT in evaluating bone lesions is its ability to measure relative density. Lesions that contain fat or are extremely dense are often benign [18, 20]. However, there is more than one way to assess lesion density (e.g., mean density, maximum density), and these values may vary depending on where the region of interest (ROI) is placed within the lesion (Fig. 8). Bone density measurements can vary slightly from scanner to scanner, with an accepted precision of approximately 3–5%, which is improved with the use of dual energy scanners over single photon scanners [37]. Intravenous iodinated contrast can elevate lesion density and even simulate a bone lesion [38], so values should be assessed on non-contrast images when possible.
Solitary lesions on MRI
High T1 lesion
MRI can detect incidental bone lesions which are occult or poorly seen on radiographs or CT. When evaluating a bone lesion on MRI, one should begin with the T1 signal characteristics. If a T1 sequence is not available, the patient should return for dedicated T1-weighted imaging
T1 signal much higher than skeletal muscle or adjacent intervertebral disc
Lesions with internal high T1 signal that matches subcutaneous fat signal on all pulse sequences are almost always benign (Bone-RADS1) [20]. The differential considerations are focal fatty marrow, intraosseous hemangioma, lipoma, osteonecrosis, Paget disease, Modic type 2 discogenic degenerative endplate changes, and other post-inflammatory focal marrow atrophy [39, 40]. When high T1 signal in a lesion is not the same as subcutaneous fat signal on all pulse sequences, both benign and malignant lesions are considerations. Here, post-contrast imaging can then help further narrow the differential diagnosis. If a T1 hyperintense lesion shows no or only thin peripheral enhancement, an intraosseous ganglion or subchondral cyst containing proteinaceous material should be considered, particularly when in a periarticular location (Bone-RADS1). Conversely, a T1 hyperintense lesion with central or mass-like enhancement would warrant a Bone-RADS4 designation, with differential considerations including melanoma metastasis or hemorrhagic metastasis. Of note, internal enhancement within intraosseous ganglia or subchondral cysts can be related to internal fibrous component, enhancing synovial joint fluid entering the cyst, or contrast diffusion into the cyst from surrounding bone marrow [41]. Melanoma metastases can present as well-circumscribed high T1 lesions owing to the paramagnetic effect of melanin and presence of hemorrhage and show solid enhancement on post-contrast images [42]. If post-contrast images are not available, a high T1 lesion not matching the subcutaneous fat signal is considered a Bone-RADS2 lesion, and additional imaging with contrast or chemical shift sequences is recommended for further characterization. On chemical shift images, a signal drop of > 20% supports a benign diagnosis, while a signal drop of less than or equal to 20% is indeterminate and cannot differentiate between benign and malignant neoplasms. Using 20% signal drop as a cutoff has a 95–100% sensitivity and 61–95% specificity[43,44,45,46,47].
T1 signal slightly higher than skeletal muscle or adjacent intervertebral disc
Focal islands of red marrow are the classic lesion in this category and are common diagnostic dilemmas. One should search for the presence of macroscopic internal fat in the lesion which is highly predictive of benignity [20]. Red marrow can have concerning MRI appearance as some lesions can be mass-like and may not show evidence of macroscopic fat. For these equivocal areas of red marrow, chemical shift imaging with in-phase and out-of-phase T1-weighted gradient images may be helpful in differentiating between red marrow from neoplastic processes [48]. If chemical shift imaging is not available, these slightly high T1 lesions are considered Bone-RADS2, and additional chemical shift imaging, radiographs, and/or bone scan should be performed. The red marrow lesion would be normal on plain film radiograph, and the bone scan would not show uptake. A focal marrow abnormality that fails to show macroscopic or microscopic fat mandates further investigation, and we recommend applying the “low T1” algorithm to these lesions.
High T1 signal in fluid level or due to hemorrhage
Hemorrhagic lesions with T1 bright areas can mask the underlying lesions and prevent optimal assessment and deserve special mention (Fig. 9). We recommend following the “High T2” algorithm for lesions with intralesional high T1 signal attributed to fluid–fluid or fluid-blood level, as these lesions most likely have simultaneous high T2 signal, and the hemorrhage could be masking an underlying lesion.
Low T1 lesion
If lesion shows low T1 signal, the next step is to evaluate T2 signal characteristics.
Low T1 high T2 lesions
Many neoplastic diseases are characterized by low to intermediate signal intensity on T1 and high signal intensity on T2 when compared to normal muscle [39, 49]. When detected, both primary benign and malignant bone tumors as well as metastatic lesions should be considered [50, 51]. Although MRI cannot accurately differentiate between the benign and malignant lesions solely based on signal characteristics, the incorporation of other imaging features can help in the diagnostic management of these lesions.
Presence of concerning imaging features
When faced with a low T1 and high T2 incidental bone lesion, a search for concerning imaging features should be a primary focus. Presence of any of the following concerning clinical or imaging findings would result in a Bone-RADS4 designation: history of malignancy with propensity for bone metastases, pain attributable to the lesion, cortical involvement, soft tissue extension, pathologic fracture, surrounding bone marrow edema, solid mass-like enhancement, a lesion in the sternum in a patient with breast cancer (Fig. 10), or elevated prostate-specific antigen (PSA) [52].
Absence of concerning imaging features
If a low T1 and high T2 lesion does not have any of the above-listed concerning features, it should be assessed for characteristic imaging features of common benign lesions (Bone-RADS1). Enchondromas, subchondral cysts, osteochondromas, NOF, and fibrous dysplasia are representative lesions. Finally, for those low T1 and high T2 lesions without concerning features and but are not consistent with the listed common bone lesions, additional imaging evaluation (Bone-RADS2) with CT to look for internal matrix, or follow-up imaging (Bone-RADS3) with repeat MRI to confirm stability, should be considered.
Low T1 and low T2 lesions
One of the main considerations for a low T1 and low T2 lesion is an enostosis; however, breast and prostate carcinoma can have purely osteoblastic metastases with corresponding low T1 and T2 signal on MRI. To distinguish between an enostosis and an osteoblastic metastasis, these lesions should be assessed for the concerning features: halo sign, solid mass-like enhancement, sternal lesion in a patient with breast cancer, or elevated PSA. The “halo sign” (Fig. 10d) is defined as a rim of abnormal increased T2 signal in the marrow surrounding the lesion of interest and has a reported sensitivity of 75%, specificity of 99%, and accuracy of 88% of being a metastasis [20]. In other words, for lesions that have the “halo sign,” 75% are metastases. Conversely, the “halo sign” is very rare in benign lesions, and 99% of benign lesions do not have the “halo sign.” Therefore, its presence should prompt further workup and referral (Bone-RADS4) [20]. Intralesional solid enhancement following administration of contrast is also a concerning feature that can be seen in osteoblastic metastases [53, 54]. A solitary sternal lesion in a patient with known breast cancer has a 76% likelihood of being metastatic [55]. For this reason, these lesions should receive a Bone-RADS4 designation.
A low T1 and T2 lesion with no concerning features in patients with history of malignancy with propensity for bone metastases may require further imaging or imaging follow-up. Options include MRI with contrast to evaluate for possible enhancement, CT to measure the HU of the lesion (refer to CT sclerotic bone lesion), or bone scan. Further workup may depend on the nature of underlying malignancy; for example, probability of bone metastasis in patient with history of prostate cancer can be predicted based on the clinical staging, prostate-specific antigen (PSA) level, and Gleason score [56]. If there are no concerning imaging features and history of malignancy, a low T1 and T2 lesion is most likely a benign lesion (Bone-RADS1), such as an enostosis, or an involuted NOF.