Magnetic Resonance Imaging (MRI) provides excellent spatial and contrast resolution without ionizing radiation, making it a favorable imaging modality during pregnancy [1]. However, due to certain concerns, MRI may be avoided as it incites anxiety among patients and families [2]. In this review, we address MRI safety concerns in pregnancy with supportive data from animal and human studies. With this review, patients, physicians, and healthcare personnel have an up-to-date resource explaining theoretical and practical aspects of MRI safety in pregnancy.

Energy deposition

Radiofrequency (RF) pulses, used for MR signal generation, consist of oscillating electromagnetic fields [3]. Absorbed RF energy causes tissue heating, a major safety concern. The RF energy is quantified using the specific absorption ratio (SAR) and is measured in watts per kilogram (W/kg; [4]). Simulation models have shown that fetal temperature can be elevated to >38 °C after RF exposure with a whole-body SAR of 2 W/kg for >7.5 min [1, 5]. If fetal temperature is elevated >2 °C, RF thermogenesis may be teratogenic [6]. However, due to technical limitations, maternal SAR has been used as a surrogate for fetal temperature increase [2]. Maternal temperature increase of >2–2.5 °C for 30–60 min may be teratogenic; hence the Food and Drug Administration recommends a whole-body SAR of < 4 W/kg, translating to a maternal temperature increase of 0.6 °C, for 30 min [4].

The SAR depends on the spatial location of tissue in the scanner. Heating is maximum at the maternal surface and negligible at the core. Some models show that 40–70% of maternal SAR is transmitted to the fetus [7]. Current MR scanners have built-in programs based on the National Electrical Manufacturers Association recommendations that automatically estimate SAR [8]. To limit tissue heating, the International Electrotechnical Commission recommends an SAR limit of 2 W/kg in pregnant patients, amounting to a potential 0.5 °C body temperature rise [9, 10].

The SAR also depends on the pulse sequence; fast spin-echo (FSE) sequences have a higher SAR than gradient-based sequences [2, 9, 11]. Sequences using long RF trains such as single-shot fast-spin echo (SSFSE) can have higher SAR values, yet are used for fetal MRI as RF is of short duration and distributed throughout the examination, usually limiting significant temperature increase [12, 13].

SAR at a field strength of 1.5 Tesla (T)

First trimester

During the first trimester, fetal cells undergo events related to implantation and organogenesis [4]. The major concern in this phase is altered organogenesis and premature termination [4, 14]. Studies of MRI-exposed mammalian stem cells have shown alteration in these events, but there are no supporting human studies [4]. In a series of 15 first-trimester pregnancies inadvertently imaged with MRI (mean gestational age [GA]: 3.8 weeks), there were no adverse outcomes [15]. In a large population-based study, Ray et al. reported that the incidence of congenital anomalies was not significantly higher in an MRI-exposed group (33.8 per 1000 person-years vs. 24.0 per 1000 person-years in the unexposed group). The incidence of congenital anomalies was also not different based on the timing of exposure during the first trimester [1].

Second and third trimesters

In a longitudinal study on the functional outcomes of 72 fetuses exposed to MRI during the second and third trimesters (mean GA: 30.5 ± 3.1 weeks), none were found to have adverse effects as all children (mean age: 24.5 ± 6.7 months) showed age-appropriate social and motor skill development [16]. In this study, the average MRI duration was 35.2 min, with RF exposure accounting for 44.5% of the duration [16]. Zvi et al. reported no adverse neurodevelopmental outcomes in pregnancies exposed to 1.5 T where they included 8, 28, and 95 patients exposed to MRI during the first, second, and third trimesters, respectively [17].

SAR at a field strength of 3 T

There is a fourfold increase in SAR in 3‑T systems compared with 1.5 T, resulting in more tissue heating. Cannie et al. demonstrated an increase of <1 °C in fetal pigs if the MRI was limited to < 30 min, but temperature increases of > 2.5 °C if MRI was performed using high SAR sequences for >1 h [18].

There is limited literature on the effects of 3‑T exposure on fetuses in the first trimester [2]. Chartier et al. found no difference in birth weight between 14 fetuses exposed to 3‑T MRI during the first trimester and unexposed controls [19]. A study of 12 pregnant patients exposed to T2-SSFSE showed that decreasing flip angle and increasing repetition time lowered the SAR without compromising the signal-to-noise ratio at 3 T [20]. In fact, these modifications lowered SAR below 25% of the 2‑W/kg limit [20]. Barrera et al. found that the mean SAR was equivalent for 1.5 T and 3 T (1.09 ± 0.69 W/kg vs. 1.14 ± 0.61 W/kg) in 2952 pregnant women with a mean gestational age of 24 weeks ±6 (range: 17–40 weeks; [12]). Another issue of MRI at a field strength of 3 T is the focal hot spot caused by RF inhomogeneity and standing waves that increase SAR [21]; this can be mitigated by RF shimming and multichannel independent transmit coils that tailor RF [22].

Acoustic noise

During MRI, Lorentz forces generated due to alternating gradient coil currents [23] cause vibration of coils resulting in acoustic noise [24]. These forces are proportional to the strength of the magnet and gradient currents [2]. In practice, acoustic noise levels can exceed 110–130 A-weighted decibel (dBA; [19, 25, 26]). Noise also depends on the spatial relation with the scanner (louder at the entrance of the bore) and hardware design [25]. Noise increases with higher duty cycles, reduced slice thickness, narrowing field of view, decreased repetition and echo times, and use of fast techniques [24]. The RF pulses also generate noise; however, this is negligible and obscured by the louder gradient current-induced noise [23, 24].

The concern related to noise is potential damage to the fetal hearing organs. In a study of behavioral responsiveness to auditory stimuli, it was reported that fetuses respond to low-frequency (100–500 Hz) high-intensity (> 100 dBA) stimuli by the 19th gestational week [27]. By the 33rd and 35th gestational weeks, fetuses responded to 1000 Hz and 3000 Hz low-intensity (< 100 dBA) tones, respectively [27]. Histopathological analysis of the organ of Corti in fetal lambs exposed to noise of 120 dB for 16 h revealed damage to hair cells in the apical and middle cochlear turns [28]. However, human hearing loss is thought to be due to damage to hair cells in the basal turn [29]. Another proposed mechanism of noise-induced fetal hearing loss is malformation of the auditory apparatus, although this is not widely accepted [29]. Jaimes et al. suggest that fetal MRI examinations performed toward the end of the second and throughout the third trimester are unlikely to cause adverse morphological effects on the fetal hearing apparatus [29].

Noise is attenuated by maternal tissues and amniotic fluid. Studies using a fluid-filled abdominal model in a 0.5‑T scanner have shown noise attenuation of 30 dB [30] with frequencies of < 0.5 kHz penetrate the uterus more easily than higher frequencies [31]. Gerhardt et al. demonstrated progressive sound attenuation of 35.3 dB for 0.5 kHz and 45.0 dB for 2.0 kHz [32]. Other important considerations are the fact that prenatal sound conduction is primarily through bone, which attenuates high frequencies, and also that hearing damage in animal studies occurs after prolonged exposure compared to the 30–45-min duration of clinical MRI [29].

Noise at a field strength of 1.5 T

First trimester

In a study evaluating exposure to a field strength of 1.5 T in pregnancy, Ray et al. followed up 1737 fetuses exposed to MRI in the first trimester until the age of 4 years. There was no significant difference in the risk of hearing loss between the study and control groups (hazard ratio: 1.24; 95% CI: 0.94–1.63; [1]). In a similar study evaluating 44 neonates exposed to in utero MRI, Strizek et al. found no adverse auditory effects [33].

Second and third trimester

In the same study by Strizek et al., 96 fetuses imaged during the second and third trimester showed no hearing deficits [33]. Reeves et al. also reported similar findings [34]. Bouyssi-Kobar et al. found that all newborns in their study of 72 pregnancies (mean GA: 30.5 ± 3.1 weeks) passed the transient evoked otoacoustic emission (TEOAE) test and had normal pre-school age hearing function [16]. Kok et al. also reported no hearing loss among 41 children exposed to MRI during the third trimester [35].

Noise at a field strength of 3 T

Chartier et al. evaluated 81 neonates exposed to prenatal MRI at a field strength of 3 T within the first 12 h of postnatal life and observed no difference in the incidence of auditory impairment between the exposed and unexposed groups (p = 0.55; [19]). Jaimes et al. evaluated hearing loss in 62 fetuses exposed to 3 T (median GA: 30 weeks, 1 day) compared to 62 fetuses exposed to 1.5 T (median GA: 20 weeks, 2 days) and found no significant differences (p = 0.74 for TEOAE, p = 0.8) for auditory brainstem response (ABR; [29]).

Gadolinium-based contrast agents

Gadolinium-based contrast agents (GBCAs) contain a paramagnetic gadolinium molecule that is toxic in its ionized form; hence it is chelated by a ligand for use in clinical MRI [36]. The main concerns of GBCA use in pregnancy are: (1) fetal exposure due to the GBCAs crossing the placenta [37,38,39]; (2) gadolinium retention, although this is still being investigated [2]; and (3) nephrogenic systemic fibrosis (NSF; [11, 40, 41]).

GBCAs in the first trimester

Teratogenic or mutagenic effects following fetal GBCA exposure are thought to occur during organogenesis, i.e., the first trimester, which is the period with the highest risk of inadvertent GBCA administration [1, 42]. Effects of 2.0 mmol/kg/day gadobenate dimeglumine administration in pregnant rabbits showed a higher incidence of intrauterine deaths, microphthalmia, retinal irregularities, thoracolumbar fusion abnormalities, and increased rates of sternal ossification center development [43]. In a study of pregnant mice, two groups received intraperitoneal injection of gadopentetate dimeglumine at 9.5 days of gestation. Of these, one group was not exposed to MRI. There was no increased incidence of abortions, stillbirths, or morphological anomalies [44].

Human studies with very limited sample sizes described no teratogenesis with GBCA exposure in the first trimester [45]. The offspring of 26 women exposed to gadopentetate dimeglumine during the periconceptional period and first trimester were found to have no neonatal complications [46]. In another study of 13 pregnancies with GBCA exposure in the first trimester, no GBCA-related malformation was reported [47].

GBCAs in the second and third trimesters

Animal studies show that GBCAs cross the placenta into the fetal circulation in the second and third trimesters, with rapid fetal clearance [2]. In a study to establish maternofetal pharmacokinetics of gadoterate meglumine in mice, Mühler et al. found the greatest fetal concentration (0.08% of injected dose) 30 min after administration, which became undetectable after 48 h. They concluded that it was redistributed back to the maternal circulation [48]. Okazaki et al. injected 0.3 mmol/kg 14C-labeled gadodiamide into rats on the 18th day of gestation and found that the maximum placental concentration was at the end of the injection (18–30% of dose), with a 100-fold decrease by 24 h [37].

Among the few human studies, Marcos et al. confirmed no neonatal adverse effects following a single dose of 0.1 mmol/kg of gadopentetate dimeglumine in 11 pregnant women (GA: 16–37 weeks; [49]). Tanaka et al. injected six pregnant patients with 5 mmol of gadopentetate dimeglumine (GA: 34–38 weeks); of the neonates born from these pregnancies, one had jaundice and one had meconium-related complications. The causal relation was unknown; all neonates were discharged without sequalae [50]. In two other studies using 0.1 mmol/kg gadodiamide and gadopentetate dimeglumine in a cohort of 29-week (mean GA: 27 weeks) and 11-week (GA: between 19 and 34 weeks) pregnancies, no adverse neonatal outcomes were observed [51, 52].

Winterstein et al. conducted a retrospective study evaluating the risk of fetal and neonatal death as well as the rate of neonatal intensive care unit (NICU) admission in patients exposed to GBCAs during pregnancy. Among the 5991 included pregnancies, the incidence of fetal/neonatal deaths was 1.4% in both the exposed and unexposed groups. The adjusted relative risk was 0.73 (95% CI: 0.34–1.55) while the adjusted relative risk for NICU admission was 1.03 (95% CI: 0.76–1.39). Overall, while their findings suggest no fatal or severe acute effects related to GBCA exposure, subacute complications were not evaluated [53].

Fetal GBCA retention

After intravenous gadoteridol administration in gravid macaques, low gadolinium levels were found in juvenile tissues, mostly in the femur and liver (mean percentage injected dose per gram of tissue of 2.5 × 10−5%ID/g and 0.15 × 10−5%ID/g, respectively; [45]). Novak et al. injected 0.1 mmol/kg of gadopentetate dimeglumine in gravid rabbits and found that the GBCA concentration in the placenta was highest at 5 min post-injection but decreased biexponentially by 60 min. Fetal tissues retained low gadolinium concentrations except the kidneys, where concentration increased with time (4.3 ± 1.1 μg/g at 5 min to 6.8 ± 1.8 μg/g at 60 min; [54]).

A potential risk of gadolinium retention is NSF in childhood. Ray et al. investigated an NSF-like outcome in children up to the age of 4 years with in utero GBCA exposure and found the incidence was higher in the exposed group (3.3 per 1000 person-years) than in the non-MRI group (1.8 per 1000 person-years) with a wide confidence interval for adjusted hazard ratio (aHR: 1.00, 95% CI: 0.33–3.02) and adjusted risk difference (0.0, 95% CI: −2.2–6.7). The incidence of adverse rheumatological, inflammatory, or infiltrative dermatological outcomes were higher following GBCA-enhanced MRI (125.8 per 1000 person-years) than in those not exposed to MRI (93.7 per 1000 person-years) groups. Risk was higher with first-trimester exposure (aHR: 1.41; 95% CI: 1.11–1.79; [1]).

Due to the aforementioned concerns, exposure to GBCAs during pregnancy should usually be avoided whenever feasible.

GBCAs and breastfeeding

The recommended pediatric dose of GBCAs is 0.1–0.2 mmol/kg, which is well tolerated up to 6 months of age [41]. Studies have reported that GBCAs demonstrate minimal milk and plasma protein binding, which minimizes its excretion in breast milk [55]. In a study by Shellock et al., 20 lactating women were administered 0.1 mmol/kg gadopentetate dimeglumine (19/20) and one lactating woman was administered 0.2 mmol/kg. The authors found that <0.04% (0.57 ± 0.71 mmol) of the maternal dose of GBCA passed into breast milk. Extrapolating from this, the amount ingested by a nursing infant is estimated to be < 1% of the permitted dose [56]. In reports by Schmiedl et al. and Rofsky et al., 0.1 mmol/kg of gadopentetate dimeglumine administered to lactating women resulted in 0.011% and 0.023% of the dose being excreted via breast milk, respectively [57, 58].

Medications to reduce fetal motion

Fetal movements begin by the seventh to eighth week of gestation with strong, sudden movements occurring around the 13–16th week [59, 60]. Multiple methods have been attempted to reduce fetal motion and improve the diagnostic quality of MRI [61]. In a survey of 67 European institutions, 24% used some form of maternal sedation (antihistamines, benzodiazepines) to reduce fetal motion [62]. Meyers et al. administered diazepam in 19 pregnant patients undergoing fetal MRI and found no effect on fetal motion [63]. Some authors suggest maternal fasting or avoiding caffeine 4 h prior to the MRI [64, 65]. Yen et al. found no association between any food, beverage, or macronutrient and fetal motion [66].

Technical advances have led to use of faster sequences, including the use of parallel-imaging techniques, which eliminate the need for sedatives [2]. By far one of the most commonly used sequence is SSFSE [67]. T1-weighted images are acquired using two-dimensional gradient echo (2D GRE) sequences and/or ultrafast gradient echo sequences or turbo fast low-angle shot (FLASH; [67]). Since SSFSE uses single-slice acquisition, fetal motion usually affects only the slice being acquired. In gradient echo sequences all slices are obtained simultaneously; thus, fetal motion degrades the entire dataset [67]. Currently, there are no globally accepted guidelines for the use of medications to reduce fetal motion.

Occupational exposure

The potential effects of RF energy deposition and acoustic noise extend to pregnant MRI personnel working around an MR field [2]. These may include technologists, physicians, nurses, physicists, and receptionists [2]. However, there are some differences compared to pregnant patients:

  1. 1.

    The duration of electromagnetic field exposure. The major proportion of duties carried out by personnel are limited to American College of Radiology (ACR) Zones I, II, and III [68]. Time spent in ACR Zone IV (scanning room) is limited, implying a short duration of RF pulse, gradient, and static field exposure [68]. Outside Zone IV, RF and gradient fields sharply decline, although there may still be some exposure to static magnetic fields [69, 70].

  2. 2.

    Exposure to electromagnetic fields over the duration of pregnancy in pregnant personnel [2].

  3. 3.

    Exposure in pregnant personnel carries no benefit to the fetus as opposed to a pregnant patient, where benefits usually outweigh risks [2].

Electromagnetic field exposure

While performing an MRI examination, personnel may be exposed to time-varying magnetic fields, static magnetic fields, and pulsed RF fields [2]. The International Commission on Non-Ionizing Radiation Protection suggests that these are unlikely to affect heart rate and rhythm below a field strength of 8 T [71]. In a review of the safety of static magnetic fields, Schenck mentions that results on humans and animals in fields up to 8 T and 16 T, respectively, indicate a substantial safety margin that is far above the field strengths of clinical MRI scanners typically used for fetal MRI [72].

Kanal et al. obtained data from 280 pregnant MRI personnel and 1141 pregnant non-MRI personnel. They found no correlation of working in an MRI environment and adverse outcomes, e.g., spontaneous abortions, infertility, premature delivery, or low birth weight [73]. There are no data available addressing the outcomes in pregnant personnel working in a specific field strength or the risks at different trimesters [2]. International societies have deemed it safe for pregnant MRI personnel to work around an MRI environment, provided they are not present in Zone IV during image acquisition [2].

Acoustic noise

Griffiths et al. recorded ABR from fetal sheep before and after exposure to 120 dB for 16 h and found prolonged latency-intensity functions with temporary elevation of ABR thresholds by an average of 8 dB [74]. When pregnant guinea pigs, in the final trimester, were exposed to 115 dBA for 7.5 h/day, it was observed that offspring demonstrated longer peak level IV latencies. Overall, these results suggest noise-induced hearing loss may occur even after inner ear development [75].

Guven et al. compared TEOAE and ABR tests between 65 infants with mothers exposed to noise (80–85 dBA for 8 h/day) during pregnancy (mean GA: 32.58 ± 2.71 weeks) to 2588 infants with mothers without noise exposure [76]. They found no difference in test results between the groups. Lalande et al. examined 131 children born from mothers that worked in noisy conditions ranging from 65 to 95 dBA, 8 h a day during pregnancy. They found a threefold increase in the risk of developing high-frequency hearing loss among children exposed to 85–95 dB over 9 months of pregnancy [77]. Selander et al. reported a slightly higher risk of neonatal hearing dysfunction in those exposed to < 84 dBA for 8 h (aHR: 1.05), and greater risk when exposure was >85 dBA (aHR: 1.27). Longer exposure (full-time employment vs. part-time employment vs non-working) and less absenteeism (< 20 days) was associated with a higher chance of hearing dysfunction (aHR: 1.82; [78]). The level and length of noise exposure in these studies far exceeded the occupational exposure to MRI noise.

A summary of the concerns and recommendations regarding MRI exposure in pregnancy is provided in Table 1.

Table 1 Magnetic resonance imaging exposure: concerns and recommendations
Full size table

Discussion

The decision to perform fetal MRI requires a thoughtful clinical approach. Needless to say, any test added to the regular array of investigations will cause maternal, family, and healthcare team anxiety [79]. According to the Canadian Association of Radiologists (CAR) expert panel, use of 1.5 T is considered safe to perform during any trimester. In fact, they suggest the use of 1.5 T over 3 T, when both are available, due to more supporting safety literature [2]. Regardless, of the field strength, MRI must always be performed in the normal operating mode [2]. With respect to first-trimester scans, delaying the MRI examination could be considered, if feasible, as fetal MRI itself is limited by the small fetal size and rapid fetal movements [2, 80]; this of course does not apply to MRI examinations performed for maternal indications.

When planning to perform an MRI examination on a pregnant patient, informed consent not only pertains to willingness to enter an MRI scanner but usually involves a detailed conversation regarding its necessity and diagnostic value [81]. Such conversations may alleviate anxiety and facilitate patient decision-making. The referring physicians and radiologists can work together in providing information in advance allowing time for the patient to clarify queries [82].

From our literature review, GBCAs in pregnancy are not recommended unless absolutely necessary. Any consideration of its use would entail critical scrutiny of the risk-versus-benefit assessment, multidisciplinary discussions, and documentation of informed consent. Following maternal GBCA administration in the postnatal period, breastfeeding has been considered to be safe [2]. Breast milk pumped within 24 h of a GBCA-enhanced MRI examination may be discarded as per the choice of the mother; however, there is no value in discarding breast milk beyond 24 h of GBCA administration [2].

Conclusion

Magnetic resonance imaging is an important tool in fetal medicine. New developments are anticipated to further improve detection and classification of congenital anomalies. Animal studies have enhanced our understanding of potential maternofetal effects of MRI. While some adverse effects have been reported in animals, it must be acknowledged that these were often described following prolonged MR exposure, which would be unreasonable in clinical practice. Most human studies show unenhanced MRI to be beneficial without significant adverse outcomes during or after pregnancy. Moreover, MRI exposure in pregnant personnel is also safe, provided precautions are exercised. Overall, judicious use of MRI and being abreast of research are suggested for the most optimal advantages of this modality.