Placental magnetic resonance imaging T2* measurements in normal pregnancies and in those complicated by fetal growth restriction
ABSTRACT
Objectives
The magnetic resonance imaging (MRI) variable transverse relaxation time (T2*) depends on multiple factors, one important one being the presence of deoxyhemoglobin. We aimed to describe placental T2* measurements in normal pregnancies and in those with fetal growth restriction (FGR).
Methods
We included 24 normal pregnancies at 24–40 weeks' gestation and four FGR cases with an estimated fetal weight below the 1st centile. Prior to MRI, an ultrasound examination, including Doppler flow measurements, was performed. The T2* value was calculated using a gradient echo MRI sequence with readout at 16 different echo times. In normal pregnancies, repeat T2* measurements were performed and interobserver reproducibility was assessed in order to estimate the reproducibility of the method. Placental histological examination was performed in the FGR cases.
Results
The method was robust regarding the technical and interobserver reproducibility. However, some slice-to-slice variation existed owing to the heterogeneous nature of the normal placenta. We therefore based T2* estimations on the average of two slices from each placenta. In normal pregnancies, the placental T2* value decreased significantly with increasing gestational age, with mean ± SD values of 120 ± 17 ms at 24 weeks' gestation, 84 ± 16 ms at 32 weeks and 47 ± 17 ms at 40 weeks. Three FGR cases had abnormal Doppler flow, histological signs of maternal hypoperfusion and a reduced T2* value (Z-score < –3.5). In the fourth FGR case, Doppler flow, placental histology and T2* value (Z-score, −0.34) were normal.
Conclusions
The established reference values for placental T2* may be clinically useful, as T2* values were significantly lower in FGR cases with histological signs of maternal hypoperfusion. Copyright © 2015 ISUOG. Published by John Wiley & Sons Ltd.
INTRODUCTION
Fetal growth restriction (FGR) due to placental dysfunction is associated with reduced oxygen supply to the fetus. Fetal oxygen supply can be assessed directly only by analysis of fetal blood obtained by cordocentesis1. However, this invasive procedure is associated with a non-negligible risk of fetal loss. Therefore, a non-invasive method for the direct assessment of placental oxygenation is needed as a supplement to current methods for diagnosis and surveillance of placental dysfunction, such as Doppler flow measurements and biophysical profile. These methods estimate fetal wellbeing, and estimate only indirectly placental function2.
Magnetic resonance imaging (MRI) can be used for the assessment of placental oxygenation, as the transverse relaxation time (T2*) depends on the presence of deoxyhemoglobin. The paramagnetic properties of deoxyhemoglobin affect the spin of neighboring protons, thereby creating magnetic field inhomogeneities that decrease the T2* value3; thus, in hypoxic tissues T2* is reduced. Apart from tissue oxygenation the T2* value is also affected by elements of tissue composition (the intrinsic T2 value) such as cell density, water content, amount of hydrogen atoms and surface area4.
In pregnancies complicated by FGR, the placenta is considered to be hypoxic mainly because of impaired transformation of the spiral arteries into low-resistance vessels, leading to reduced maternal perfusion of the placenta5, 6. Under these circumstances, the T2* value should be reduced in the placentae of FGR pregnancies and we therefore hypothesize that placental T2* value is a sensitive marker of placental dysfunction.
In this study, we describe placental T2* values in normal pregnancies and their dependency on gestational age, and we investigate the reproducibility of the measurement of T2*. In addition, four FGR cases were assessed to investigate the association between placental T2* measurements, Doppler flow in fetal and uterine arteries and placental histology. To our knowledge, this is the first study to describe placental T2* values in human FGR pregnancies.
METHODS
Subjects
We included 24 normal pregnancies at 24–40 weeks' gestation, all with normal ultrasound-based estimated fetal weight (EFW)7 and normal Doppler flow measurements in the umbilical artery, fetal middle cerebral artery and maternal uterine arteries. We also included four FGR cases with an EFW < 1st centile. In three of the four cases Doppler flow was abnormal. In the remaining FGR case, the Doppler findings were normal (Table 1). Case 1 was diagnosed with antiphospholipid syndrome and lupus anticoagulant prior to pregnancy. Case 2 was a smoker, and thrombophilia screening 3 months postpartum revealed Factor V Leiden thrombophilia. Case 3 had mild pre-eclampsia and Case 4 was healthy. The study was approved by the Regional Committees on Biomedical Research Ethics (Journal number M-20090006 and N-20090052). Oral and written informed consent was obtained from all participating women.
| Case | GA at MRI (weeks) | EFW (g) | Doppler findings (Z-score) | Placental T2* (ms) | Outcome | BW (g) | Placental histological examination |
|---|---|---|---|---|---|---|---|
| 1 | 24 + 5 | 361 (−51%†) |
CPR: 0.40 (−4.26) DV-PI: 1.63 (6.85) UtA-PI: 1.69 (1.89) |
27 | Stillbirth, vaginal delivery at 26 + 0 weeks | 380 |
Low weight: 93 g (−59%‡) Normal shape Decidual arteriopathy (atherosis) Accelerated maturation Distal villous hypoplasia Increased intervillous fibrin deposition Multiple infarcts (65%) |
| 2 | 29 + 0 | 962 (−31%†) |
CPR: 0.58 (−4.01) DV-PI: 0.77 (1.36) UtA-PI: 1.90 (2.58) |
39 | Acute CS at 29 + 2 weeks, good outcome | 974 |
Low weight: 185 g (−37%‡) Abnormal shape (‘pancake’) Accelerated maturation Multiple infarcts (50%) |
| 3 | 30 + 0 | 956 (−43%†) |
CPR: 0.89 (−3.28) DV-PI: 0.73 (1.21) UtA-PI: 2.37 (3.43) |
30 | Acute CS at 31 + 0 weeks, good outcome | 1050 |
Low weight: 191 g (−44%‡) Abnormal shape (‘cupcake’) Accelerated maturation Few infarcts (5%) Marginal cord insertion Fetal thrombotic vasculopathy |
| 4 | 31 + 1 | 1094 (−39%†) |
CPR: 1.44 (−1.85) DV-PI: 0.54 (0.01) UtA-PI: 0.77 (−0.54) |
77 | Elective CS at 34 + 0 weeks, good outcome | 1515 |
Low weight: 267 g (−35%‡) Normal shape Normal histology |
- † Relative to mean estimated fetal weight (EFW) in normal pregnancy7.
- ‡ Relative to mean placental weight in normal pregnancy10.
- BW, birth weight; CPR, cerebroplacental ratio; CS, Cesarean section; DV, ductus venosus; GA, gestational age; PI, pulsatility index; UtA, uterine artery.
MRI
All measurements were performed on a GE Discovery MR450 1.5-Tesla MRI System (GE Healthcare, Milwaukee, WI, USA). During the MRI scan, the women were placed in the left lateral position and an eight-channel cardiac coil was placed over the abdomen, covering the entire uterus. Initially, a T2-weighted localizer was performed to obtain the anatomic orientation of the placenta. This was followed by a placental T2* scan using a gradient-recalled echo sequence with the following parameters: repetition time, 70.9 ms; 16 echoes ranging from 3.0 to 67.5 ms in steps of 4.3 ms; field of view, 350 × 350 mm; and matrix, 256 × 128 mm. The size of the matrix resulted in an in-plane resolution of 1.37 × 2.73 mm. Two 8-mm slices, with a slice gap of 2 cm, were placed in the central part of the placenta in transverse orientation (Figure 1). Each slice was acquired within a single breath-hold of 12 s. In the normal group, a subset of women (n = 16) underwent repeat placental T2* scanning (scan 2) after 2 min while the woman remained in the same position inside the bore. In a subset of these 16 women (n = 8), the placental T2* scan was repeated (scan 3) in a separate scan session after a 45-min interval during which the women left the MRI suite (Figure 1).
MRI analysis
Images were processed using a program developed in-house and written in MATLAB (The MathWorks Inc., Natick, MA, USA). Regions of interest (ROI) were drawn by two independent obstetricians (M.S. and A.S.). The ROI covered the entire placenta in the transverse orientation, ensuring that the outer placental borders were not crossed (Figure 1). The same ROI was used for the 16 echo time images in each T2* scan, however it was repositioned if affected by obvious artifacts including patient or fetal movements. T2* values were obtained by fitting the averaged signal within each ROI as a function of echo time. Fitting was performed using a mono-exponentially decaying function with the equilibrium magnetization (M0) and T2* as free parameters (Figure 2). A non-linear least-squares fitting algorithm was used8, and each timepoint was weighted using the SD of the ROI, thereby giving less significance to ROIs with a high SD. T2* values were calculated by averaging the fitted T2* measurement of the two separate placental slices.
Statistical analysis
For the normal pregnancies, the relationship between the placental T2* value (mean of two slices) and gestational age was estimated using ordinary linear regression, and 95% CIs and prediction intervals were calculated. In FGR cases, the standardized variables (Z-scores) of the placental T2* were estimated.
The interobserver reproducibility of the ROI drawing and the variation in T2* measurement within and between sessions were investigated using Bland–Altman plots showing 95% limits of agreement9. Measurements from placental slice 1a in normal pregnancies (Figure 1; n = 24) were used to estimate interobserver reproducibility. The reproducibility of the method was estimated using the variation in the value of T2* obtained by repeat scans of a single placental slice. As indicated in Figure 1, the reproducibility within a session was determined by comparing slices 1a and 2a, and the reproducibility between sessions was determined by comparing slices 1a and 3a. In addition, the variation between the T2* value of two parallel slices separated by a distance of 2 cm within the same placenta was estimated (slices 1a and 1b). Finally, the reproducibility within and between sessions was calculated using the mean of two parallel slices instead of a single slice in each placenta. The statistical software package Stata®11 (StataCorp LP, College Station, TX, USA) was used for data analysis.
Placental examination
The placentae were examined according to a standard protocol, which included ‘trimmed’ weight compared with normal standards10, measurement of area and thickness, and standard sections were taken for histology (two full-thickness sections from the central placental parenchyma, one section from the umbilical cord insertion with underlying placental parenchyma, two membrane rolls, one cassette with sliced membranes, two sections from the umbilical cord and sections from macroscopic abnormalities). Histological examination by light microscopy included assessment of villus maturation (type and size of villi, number of syncytial knots and number of vasculosyncytial membranes with reference to gestational age), the amount of fibrin deposition, other villus abnormalities, fetal and maternal vasculature and description of all microscopic abnormalities.
RESULTS
The reproducibility of the method is demonstrated by Bland–Altman plots in Figure 3. The 95% limits of agreement for the within- and between-session variation for a single slice were −2.1 ± 10.4 ms and −0.6 ± 22.6 ms, respectively, and for the mean value of two slices they were −1.1 ± 7.0 ms and −0.0 ± 17.8 ms. Thus, agreement within and between sessions was improved by averaging the T2* values of two slices. The agreement between measurements of two parallel slices within the same placenta was −1.1 ± 17 ms (data not shown). Based on these findings, we decided to give the T2* values as the average of two parallel slices separated by 2 cm. The interobserver reproducibility of the T2* measurement after ROI drawing by two independent observers was −0.1 ± 4.6 ms.
In the group of normal pregnancies, neonatal outcomes including birth weight were normal. In these pregnancies, the placental T2* value decreased by 4.6 ms per gestational week (R2 = 0.68; P < 0.001; Figure 4). The mean ± SD T2* values were 120 ± 17 ms at 24 weeks' gestation, 84 ± 16 ms at 32 weeks and 47 ± 17 ms at 40 weeks.
, Ordinary least-squares fit (R2 = 0.68; P < 0.001); 
, 95% CI; 
, 95% prediction interval.Outcomes of the FGR cases are presented in Table 1. In the three FGR cases with abnormal Doppler flow, placental histological examination revealed different abnormalities associated with maternal hypoperfusion, such as low placental weight, accelerated villus maturation and infarcts (65%, 50% and 5%, respectively)11. In addition, we found other abnormalities associated with FGR: Case 1 demonstrated distal villous hypoplasia, increased intervillous fibrin deposition and maternal decidual vasculopathy (acute atherosis); Case 2 demonstrated abnormal placental shape (extremely thin; ‘pancake’); Case 3 demonstrated abnormal placental shape (extremely thick; ‘cupcake’) and fetal thrombotic vasculopathy (small groups of fibrous avascular villi) as a result of cord compression caused by a marginally inserted umbilical cord. In these FGR cases, the placental T2* value was dramatically reduced: Case 1, 27 ms (Z-score, −5.24); Case 2, 39 ms (Z-score, −3.54); Case 3, 30 ms (Z-score, −3.84). Furthermore, on the T2*-weighted MRI scans these pathological placentae appeared darker than normal (Figure 5). In FGR Case 4 with normal Doppler findings, the placental histology was normal and the placental T2* value was within the normal range (82 ms; Z-score, −0.34).
DISCUSSION
In this study, we have demonstrated that placental T2* estimates are robust and the reproducibility of the method is better when based on the average of two placental slices. In normal pregnancies, the placental T2* value decreases as pregnancy progresses. Furthermore, from our FGR cases, we found that the placental T2* value was significantly reduced only in FGR cases with abnormal Doppler findings and histological signs of maternal hypoperfusion. This finding indicates that the placental T2* value might be a sensitive marker of placental dysfunction.
The exact correlation between placental T2* value and placental histological findings still needs to be described in detail. In general, T2* is closely related to T2, which depends on elements of tissue morphology. In a study by Wright et al.12 placental T2 was correlated to the amount of macromolecular deposition, such as fibrin. However, placental T2 did not correlate with villus volume density or total blood volume. In addition, the T2* value is also highly dependent on local magnetic field inhomogeneities as created by the presence of deoxyhemoglobin, for instance13. Thus, placental T2* is associated with tissue morphology as well as tissue oxygenation.
A limitation of this study is the lack of placental histological examination in the group of normal pregnancies. However, in the normal group, ultrasound Doppler flow measurements of uterine arteries, neonatal outcomes and macroscopic placental examinations were normal and therefore we assumed that the placentae in these pregnancies were normal.
We have shown that, in the normal placenta, T2* varies between parallel slices obtained within the same placenta. This finding might be explained by the heterogeneous nature of the normal placenta, as it contains fetal as well as maternal compartments. The transverse orientation of the MRI slices ensures that each slice contains both compartments. However, if the compartments are not equally represented in both slices, T2* may differ because of the different morphology and oxygenation in the two compartments. Based on these considerations, we suggest that the mean placental T2* value of the entire placenta should be calculated as the average of at least two slices.
In order to estimate the general reproducibility of the method, we investigated repeat T2* measurements and interobserver agreement. Repeat T2* measurements were performed within and between sessions for a single placental slice and for the mean T2* value of two slices. The within-session T2* variation can be explained by automatic scanner calibrations, such as adjustment of gain and frequency, and by slight displacement of the slice caused by unequal breath-holds. Our within-session agreement suggests that the method is reproducible technically. The T2* changes between sessions introduce another level of variation due to maternal physiological changes related to, for example, degree of hydration, temperature and distribution of intestinal gas, which also affect the T2* value. Furthermore, replanning of placental slices is performed between sessions and therefore it is not the same pair of placental slices being investigated in each session. Both of these factors negatively affect between-session agreement. We demonstrated that averaging measurements from two parallel placental slices improves the agreement of placental T2* measurements within and between sessions. Furthermore, interobserver agreement in our study indicates that drawing placental ROIs on the MR image can be performed easily.
From a clinical perspective it should be noted that, in our FGR cases, placental dysfunction was associated with T2* values below three SDs. Therefore, given the reproducibility of placental T2* measurements demonstrated in this paper, the method should be able to distinguish between normal and pathological placentae.
In contrast to our findings, Huen et al.14 did not find any correlation between placental T2* value and gestational age in normal pregnancies. However, our finding is supported by those of other studies on placental T2, demonstrating a decrease in the T2 value of 4.0 ms/gestational week using a 0.5-T MRI system15 and 2.4 ms/gestational week with a 1.5-T MRI system12. We suggest that the decrease in placental T2* value demonstrated in our study can be explained partly by the morphological maturation of the normal placenta and partly by the well-known decrease in placental oxygenation as pregnancy advances16. Our findings are also in line with previous findings of decreased placental perfusion with increasing gestational age, demonstrated by placental diffusion-weighted MRI17.
To our knowledge, this is the first study on placental T2* in human FGR pregnancies. However, two animal studies have been published demonstrating a reduced placental T2* value during experimental FGR18 and hypoxia19. Human placental T2 measurements have been described in a few FGR studies. In these studies, placental T2 was lower in FGR cases than in normal pregnancies20-22. Based on these studies, we suggest that the markedly reduced placental T2* values demonstrated in the pathological placentae in our study can be explained by placental hypoxia and by focal pathology, such as increased necrosis and fibrosis, which might affect the intrinsic T2 value23.
Future studies should include more FGR cases, as the small number of such cases in this study is obviously a limitation. Furthermore, in future studies the timing of the reduction in placental T2* values should be investigated. Derwig et al.22 studied placental T2 in mid-pregnancy and found that it was reduced even before clinical signs of impaired placental function became apparent. As T2* is more sensitive to hypoxia than is T2, placental T2* might be a better predictor of FGR than is placental T2.
In conclusion, placental T2* measurements have the potential to become a non-invasive test of placental dysfunction. Thus, in the small fetus, a placental T2* value below normal may be a strong indicator of placental dysfunction as the mediator of growth restriction.
ACKNOWLEDGMENTS
We thank statistician Lars Oddershede at the Institute of Clinical Medicine, Aalborg University Hospital, for fruitful discussions regarding statistical methods. Furthermore, we are grateful to radiographer Carsten Simonsen at the Department of Radiology, Aalborg University Hospital, for his expert assistance. This study was funded by Speciallæge Heinrich Kopps Legat and by Region Nordjyllands Sundhedsvidenskabelige Forskningsfond.
