Accuracy of Multimodality Fetal Imaging (US, MRI, and CT) for Congenital Musculoskeletal Anomalies

Background: Ultrasonography (US) is the first-line diagnostic tool used to assess fetal musculoskeletal (MSK) anomalies. Associated anomalies in other organ systems may benefit from evaluation via Magnetic Resonance Imaging (MRI). In this study, we compared the diagnostic accuracy of US and MRI to diagnose fetal MSK (primary objective) and non-MSK anomalies (secondary objective). We describe additional findings by low-dose computerized tomography (CT) in two cases incompletely characterized via US and MRI. Materials and Methods: This was an IRB-approved retrospective study of consecutive patients with suspected fetal MSK anomalies examined between December 2015 and June 2020. We compared individual MSK and non-MSK anomalies identified via US, MRI, and CT with postnatal outcomes. Sensitivity and specificity for US and MRI were calculated and compared. Results: A total of 31 patients with 112 MSK and 43 non-MSK anomalies were included. The sensitivity of MRI and US for MSK anomalies was not significantly different (76.6% vs. 61.3%, p = 0.3). Low-dose CT identified eight additional skeletal anomalies. MRI diagnosed a higher number of non-MSK anomalies compared to US (81.4% vs. 37.2%, p < 0.05). Conclusions: Fetal MRI and US have comparable sensitivity for MSK anomalies. In selected cases, low-dose CT may provide additional information. Fetal MRI detected a larger number of non-MSK anomalies in other organ systems compared to US. Multimodality imaging combining all the information provided by MRI, US, and CT, if necessary, ultimately achieved a sensitivity of 89.2% (95% CI: 83.4% to 95.0%) for the diagnosis of musculoskeletal anomalies and 81.4% for additional anomalies in other organs and systems.


Introduction
The incidence of fetal musculoskeletal (MSK) anomalies in pregnancy is approximately 0.4 to 0.6%, dropping to 0.024% postnatally, reflecting a high mortality rate [1]. Commonly identified fetal MSK anomalies include clubfeet, polydactyly, syndactyly, spinal deformities, limb-length discrepancies, skeletal dysplasias, and arthrogryposis [2][3][4][5][6]. Skeletal dysplasias are heritable diseases that affect bone and cartilage and occur in roughly 1/5000 births. It is of the utmost importance to properly diagnose skeletal dysplasias as early as possible in utero, as numerous are lethal [7]. Certain skeletal dysplasias carry a high risk of recurrence in future pregnancies, depending on the particular inheritance pattern [8]. For this reason, understanding the correct diagnosis can assist families in planning future pregnancies.
Educating parents on the nature of the disease, the survival chances of the fetus, subsequent development abnormalities for survivors, and future reproductive risks are essential.
US is the primary imaging modality used to assess for congenital anomalies given its low cost, safety, ease of use, and availability [2]. Previously reported sensitivities for the prenatal diagnosis of skeletal dysplasias ranged from 53% to 67.9% [9][10][11][12]. A sensitivity of 63% has been reported for the prenatal diagnosis of clubfoot using US [9]. Regarding additional limb abnormalities, Dicke et al. found that US had a sensitivity of 19.1% prenatally for polydactyly, 76.0% for abnormal hand position, 76.0% for limb reduction defects, and 81.3% for arthrogryposis [10]. Data on the accuracy of fetal MRI in diagnosing MSK anomalies is limited with no direct comparison of the diagnostic accuracy between US and MRI [13,14]. Besides the evaluation of MSK anomalies per se, fetal MRI also has the potential to provide additional information in cases of syndromic skeletal dysplasias by diagnosing unsuspected associated anomalies in other organ systems (e.g., brain, lungs, kidneys, and GI tract) [2,11,12]. More recently, low-dose fetal computerized tomography (CT) with the three-dimensional (3D) reconstruction of the fetal skeleton has emerged as an attractive imaging modality for the accurate characterization of the skeletal phenotype in skeletal dysplasias [15][16][17][18]. Prior studies have shown that both the sensitivity and specificity of low-dose CT are higher compared to US, with the only limitation being the small radiation exposure in utero (usually <5 mSv) [15]. Thus, its use is limited to situations when US and/or MRI cannot satisfactorily characterize the phenotype [15].
At our institution, we perform approximately 250 fetal imaging evaluations per year using a combination of MRI and US. Anomalies referred for fetal MRI tend to be complex, usually identified via US at the obstetrician's office and further evaluated via detailed US by maternal-fetal medicine specialists in the state of Arizona. Patients are referred for fetal MRI when the fetal phenotype has not been completely characterized. After review of the MRI images, the evaluation may be complemented by a targeted US and, in the case of skeletal anomalies, by a low-dose CT with 3D rendering of the fetal skeleton, but only in cases where the phenotype could not be characterized by fetal MRI or US.
The primary objective of this study is to determine if fetal MRI can provide additional diagnostic information compared to US for the evaluation of fetuses with a suspected MSK anomaly. The secondary objective is to determine if fetal MRI provides additional diagnostic information for anomalies not involving the MSK system in this group of fetuses (non-MSK anomalies). We describe additional findings identified with low-dose computerized tomography (CT) in cases incompletely characterized by US and MRI.

Methods
This was a retrospective IRB-approved study that included consecutive pregnancies with suspected fetal MSK anomalies referred to our institution for multimodality fetal imaging (fetal MRI, US, and low-dose CT if necessary) between December 2015 and June 2020. For each case, the mother's prenatal chart, all images, and the infant's postnatal chart were reviewed. Cases of intrauterine fetal demise without postmortem X-rays or autopsy and patients who were lost to follow up were excluded. For each case, the anomalies identified with the referring detailed prenatal US and the anomalies identified via each imaging modality performed at our institution were compared to postnatal diagnoses.
Fetal US was performed using an EPIQ Elite ultrasound system (Philips Healthcare, Bothell, WA, USA). Fetal MRI was performed using either a 3 Tesla Philips Ingenia MRI System or a 1.5 Tesla Philips Achieva MRI system (Philips Healthcare, Cambridge, MA, USA). Low-dose CT was performed using a 256-slice CT scanner (Philips 256-slice Brilliance iCT scanner, Philips Healthcare, Cambridge, MA, USA).
Most examined fetuses had more than one anomaly, and each anomaly was categorized as MSK or non-MSK. MSK anomalies were categorized as anomalies affecting the craniofacial structures, spine, clavicles, scapulae, ribs, pelvis, upper extremities, and lower extremities. Non-MSK anomalies were further categorized into cardiac, central nervous system, eye, gastrointestinal, or genitourinary anomalies. A unifying diagnosis based on phenotype was attempted and compared to a postnatal diagnosis established by postnatal clinical, imaging, and/or surgical evaluations for associations or sequences (e.g., amniotic band syndrome, caudal regression, or VACTERL), or postnatal clinical, imaging, and/or surgical and genetic testing for skeletal dysplasias or genetic anomalies (e.g., hypochondrogenesis or diastrophic dysplasia).
For statistical analysis, each individual anomaly was documented as either a truepositive or a false-positive diagnosis. This was performed for each separate modality (i.e., US, MRI, and CT) and compared to each individual anomaly diagnosed postnatally. For each system considered normal (i.e., CNS, cardiac, gastrointestinal, and genitourinary for non-MSK anomalies; cranial, facial, spine, clavicles, scapulae, ribs, pelvis, upper extremities, and lower extremities for MSK anomalies), either a true-negative or false-negative diagnosis was assigned after comparison with the postnatal outcome. Sensitivity and specificity with a 95% confidence interval (95% CI) were calculated for US and MRI and compared using McNemar's test. The added value of CT, if any, is described separately. All p-values are two-sided, and p < 0.05 was considered statistically significant. Patient demographics and clinical characteristics are reported as means ± standard deviations for continuous variables and frequencies or percentages for categorical variables. Gestational age at the time of multimodality imaging was recorded.

Results
Forty consecutive singleton pregnancies with a suspected diagnosis of one or more fetal MSK anomalies were referred to our institution during the study period. Nine pregnancies complicated by intrauterine fetal demise without postmortem X-rays or autopsy (n = 4) and neonates who were lost to follow-up (n = 5) were excluded. Patient demographics are presented in Table 1. A total of 31 patients with 111 MSK anomalies and 43 non-MSK anomalies were included in this study. There were also 222 normal MSK findings and 112 normal non-MSK findings. All 31 patients underwent a fetal MRI. Twenty-one were further evaluated by a targeted US. Low-dose CT was performed in two cases (hypochondrogenesis and disorder of glycosylation mimicking Desbuquois dysplasia) [17]. Of the 31 patients included, four patients also genetic evaluation.

Diagnostic Accuracy for MSK Anomalies
Regarding MSK anomalies, the sensitivity of the referral US compared to postnatal outcome (n = 31) was 61.3% (95% CI: 52.5% to 70.3%). The sensitivity of MRI for the same cases was 76.6% (95% CI: 68.7% to 84.5%), but the difference was not statistically significant (McNemar's test 10.7, p = 0.30) ( Table 2). In a sub-analysis restricted to cases that had matched US and MRI performed at our institution (n = 21), the sensitivity of US was 79.1% (95% CI: 70.5% to 87.7%) and the sensitivity for fetal MRI was 74.4%% (95% CI 65.2% to 83.6%), also not statistically significant (McNemar's test 0.45, p = 0.50) ( Table 3). When findings from US and MRI were combined, the sensitivity increased to 82.6% (95% CI: 74.5% to 90.6%), as US detected seven anomalies that were not identifiable by MRI ("cobrahead" appearance of the spine, mesomelic limb shortening of the upper and lower extremities in a case of diastrophic dysplasia, premature ossification center of the proximal femoral epiphysis, visualized an amniotic band that was not detectable by MRI in two fetuses, and correctly identified tibial hemimelia that could not be well visualized by MRI), whereas MRI identified three anomalies that were not identified by US (cleft palate, glossoptosis, and bell-shaped thorax).

Additional Skeletal Anomalies Diagnosed by Low-Dose CT
Eight additional osseous anomalies in two cases were identified only by low-dose CT. The first four were platyspondyly, round iliac wings with horizonatal acetabular roofs, demineralized sacrum, and metaphyseal flaring of the humeri in a fetus with hypochondrogenesis. The other four anomalies were enlarged sutures and fontanelles, coronal and sagittal clefts in the thoracolumbar spine, flat acetabula, and enlarged lesser trochanters of the femora ("sweedish key" or "monkey wrench sign") in a case of a disorder of glycosylation mimicking Desbuquois dysplasia.

Discussion
This study showed that US and MRI have comparable sensitivities for the prenatal diagnosis of MSK anomalies and that, in a population of fetuses with a skeletal anomaly, MRI may add information by the identification of previously unsuspected anomalies affecting other organs and systems. The number of cases evaluated by low-dose CT in this study is small (n = 2), but in these two cases, CT identified additional skeletal anomalies that helped achieve an accurate prenatal characterization of the phenotype in a case of hypochondrogenesis and a case of a disorder of glycosilation mimicking Desbuquois dysplasia. While CT was not used frequently, incorporating this modality can provide a more detailed assessment of the fetal skeleton when compared to both US or MRI and can, as a result, allow for a more definitive diagnosis [18].
Doray et al. have previously studied the efficacy of US to prenatally identify skeletal dysplasias [19]. There were 47 cases with skeletal dysplasia that were identified with the prenatal and postnatal diagnoses compared. Of the 47 cases, 28 (60%) had an accurate prenatal diagnosis using ultrasonography [19]. This was similar to the results found in the current study, where the referring USs had a sensitivity of 58.9% for all MSK anomalies, not limited to skeletal dysplasia. Of the remaining cases, 9 (19%) had an inaccurate diagnosis, and 10 (21%) had an imprecise diagnosis [19]. Similarly, Parilla et al. examined the prenatal accuracy of US in diagnosing skeletal dysplasia over eight years [20]. In the 31 cases examined in that study, 20 (65%) had an accurate prenatal diagnosis using US. Of note, lethality was correctly predicted in 16 out of 16 eligible cases (100%) [20]. This finding was also seen by Goncalves et al., who found that US had a sensitivity of 89% in prenatally identifying a lethal dysplasia [21]. While the overall diagnosis was not always accurate, US was able to correctly predict lethality when applicable. US is an incredibly valuable tool in prenatal imaging; however, the findings in the studies by Doray et al. and Parilla et al. in addition to this current study suggest that the use of US leaves significant room for improvement in the diagnosis of MSK anomalies and, particularly, better characterization of anomalies in other involved organs and systems.
At the moment, there is limited research regarding the accuracy of stand-alone MRI for congenital MSK anomalies. Blaicher et al. examined the utility of fetal MRI in 14 patients that were found to have skeletal dysplasia in prenatal US [22]. In ten of those cases, US was more accurate in diagnosing skeletal dysplasia than MRI. In the other four cases, each with spina bifida, MRI provided additional information that was beneficial in presurgical planning [22]. Studies that featured additional organ systems showed different results. Goncalves et al. found that when examining central nervous system (CNS) anomalies prenatally, MRI was more sensitive than both 3D US and 2D US, 88.9% compared to 66.7% and 72.2%, respectively [23]. These results for CNS anomalies were similar to the ones seen in this current study, where US had a sensitivity of 26.9% for CNS anomalies, while multimodality imaging had a sensitivity of 96.2%. The same study by Goncalves et al. showed that when MRI alone was compared to 3D US and 2D US for non-CNS anomalies, the sensitivities for each modality were similar [23]. MRI also has demonstrated utility in differentiating isolated versus complex anomalies, such as amniotic band syndrome in a case of isolated limb deficiency [24]. This added diagnostic value not only allows providers to adequately approach a child's treatment but also allows the family to fully comprehend the complexity of the congenital anomalies.
One of the limitations of this study was that it was a retrospective chart review and allowed us to exclude patients who did not qualify for this study. This was also a singleinstitution study that limited the patient population. MRI was used more frequently than US, and there were only two cases where CT was used in this study, so future studies could aim at comparing a more equal number of cases from each modality. The majority of MSK anomalies were seen in the extremities, so future studies could include patients with MSK findings localized to other parts of the body. Of the 31 patients included, only four had a follow-up genetic analysis, so there was limited correlating genetic data for many of the cases.
Diagnosing MSK anomalies and skeletal dysplasias accurately in the prenatal setting is of the utmost importance. Given the morbidity and mortality associated with certain severe skeletal dysplasias, it is essential to educate parents on the disease so they can prepare for potentially unfavorable outcomes. While US and MRI demonstrated similar diagnostic accuracy to diagnose MSK anomalies, the use of MRI provided a more accurate assessment for non-MSK anomalies. Multimodality imaging combining all the information provided by MRI, US, and CT if necessary ultimately achieved a sensitivity of 89.2% (95% CI: 83.4% to 95.0%) for the diagnosis of musculoskeletal anomalies and 81.4% for additional anomalies in other organs and systems.