Perinatal mortality: clinical value of postmortem magnetic resonance imaging compared with autopsy in routine obstetric practice


*Correspondence: M. E. Alderliesten, Department of Obstetrics and Gynecology, Academic Medical Center, Postbus 22660, 1100 DD Amsterdam, The Netherlands.


Objective To compare postmortem magnetic resonance imaging (MRI) with autopsy in perinatal deaths. To determine the acceptance and feasibility of postmortem perinatal MRI.

Design Cohort study.

Setting Large teaching hospital.

Population Fetuses and neonates from 16 weeks gestational age until 28 days after birth, stillbirths as well as intrapartum and neonatal deaths.

Methods MRI was performed prior to autopsy in a consecutive cohort of perinatal deaths after full parental consent. Agreement between MRI and autopsy was calculated. The consent rate for both examinations was recorded separately, as well as the time between the perinatal death and the MRI.

Main outcome measure Full agreement between MRI and autopsy.

Results Of 58 cases, 26 parents consented to both examinations (45%). Autopsy showed 18 major malformations, of which 10 were detected with MRI. The positive predictive value of MRI was 80% (4/5) and the negative predictive value was 65% (13/20). Additional consent for MRI was given in eight cases (14%). In 84%, the MRI could be performed within 48 hours.

Conclusions MRI is of value if autopsy is refused, but diagnostic accuracy is insufficient to recommend substitution of full autopsy. The acceptance rate of MRI only is better than that of autopsy.


Pregnancies are complicated by perinatal death in about 1%1. Establishing the cause of this perinatal death is important for the counselling of parents and in taking decisions for future pregnancies2. Several studies have confirmed the diagnostic value of perinatal autopsy in postmortem examination3,4. However, studies in both the United States and Europe indicate that as much as 40% of all perinatal deaths are not submitted for autopsy3,5,6. The reason for this high percentage of refusals is largely unknown. Landers et al.7 identified the parents' socio-economic status and the clinical diagnosis as factors related to consent to neonatal autopsy in 33 cases.

In the early nineties, the first experiences with magnetic resonance imaging (MRI) as an alternative for autopsy in postmortem examinations among adults, fetuses and neonates were published8–10.

In 1996 and 1997, Woodward et al.11 and Brookes et al.12 systematically compared MRI to autopsy in 26 and 20 cases of perinatal death, respectively. Due to the inclusion process (predominantly terminated pregnancies for congenital malformations in Woodward et al.'s study or fetuses recruited from the pathology unit Brookes et al.'s study), a very high rate of malformations was found at autopsy in these studies. The value of MRI as an alternative to autopsy in a non-selected obstetric population could not be estimated.

The objective of the present study was to establish the concordance of postmortem perinatal MRI with autopsy findings in a non-selected obstetric population. The second objective was to determine the acceptance of MRI compared with autopsy and to evaluate its implementation in routine obstetric practice.


We scheduled to study a consecutive cohort of perinatal deaths in the OLVG Hospital, a general district hospital with approximately 2000 hospital deliveries and 800 home deliveries a year. The study group consisted of fetuses and neonates, who were born from 16 weeks of gestational age and who died intrauterine, during delivery or after delivery, until 28 days after birth. Consent for autopsy was asked of all parents, as part of routine perinatal care. Independent of the consent for autopsy, explicit parental consent for an MRI was asked as part of the study protocol. After consent, MRI was performed as soon as possible. Both the MRI and the autopsy were reported by the same radiologist and pathologist, respectively, in a standardised way. All organs were reviewed and the malformations observed were described. The extremities were not examined as they are easily evaluated by physical examination and conventional X-rays. The radiologist was not supplied with clinical information and both radiologist and pathologist were blinded for each other's findings. The autopsy findings were compared with those of MRI. Afterwards, the malformations found at autopsy were classified as major or minor malformations, according to Woodward et al.11, at a clinical meeting with an obstetrician, pathologist and radiologist. The concordance between MRI and autopsy per case of perinatal death, considering only the major malformations, was calculated.

MRI was performed on a Philips Gyroscan NT 1 T with a PowerTrak 3000 Gradient system. Depending on the size of the child, a specific coil was used. Small children up to 500 g were scanned using a dedicated knee coil. Larger children were scanned with ‘wrap-around’ surface coils. The imaging protocol of the head consisted of a sagittal and transverse T1-weighted sequence (TR 500 ms, TE 18 ms, matrix 512 × 512, NSA 2, slice thickness/gap 3 mm/0.6) and a transverse T2 dual weighted Turbospin echo sequence (TR 3000, TE1 11, TE2 150, matrix 512 × 512, NSA 2, slice thickness/gap 3.0 mm/0.3). The body was scanned in a coronal and sagittal directions using a T1-weighted sequence in the first 16 cases and using a T2 Turbospin echo sequence in the following 10 cases.

All analyses were performed using SPSS 10.0 statistical program. Approval from the local medical ethics committee was obtained prior to the start of this study.


During the inclusion period, 58 perinatal deaths occurred at the OLVG Hospital.

Technically, MR images were of good quality. Maceration influences the quality of MRI, as it influences the quality of autopsy; in some cases, however, very good images could be obtained even from third degree macerated fetuses (see Fig. 1). One case was excluded from analysis because the quality of the MRI was insufficient for reporting.

Figure 1.

T2-weighted MRI image of a third degree macerated fetus, born at gestational age of 25 weeks and 4 days. Autopsy was performed, but evaluating the presence of malformations was difficult due to the degree of maceration. At MRI, no malformations were detected.

In 26 cases, a combined consent for autopsy and MRI could be obtained. The mean gestational age in the 26 cases with combined consent was 25.3 weeks (range 16–40), with a mean birthweight of 909 g (range 60–3885). There was one twin pregnancy and there were 17 male and 9 female fetuses/neonates in the study group.

A total of 18 major and 8 minor malformations were found at autopsy. Fifty-six percent (n= 10) of the major malformations were detected with MRI (see Table 1). Umbilical hernia and lung lesions were the minor malformations that were detected with MRI (see Table 2). Multiple small lung lesions were visible on MRI, where at autopsy meconium aspiration was present (lung lesions).

Table 1.  Major malformations at autopsy.
CaseMajor malformationsDetected with MRIGestational age (weeks)Birthweight (g)
 anal atresia  
12complex cor vitium29695
 mega ureters  
14hypoplastic lungs+363885
 dysplastic kidneys+  
15complex cor vitium403430
18left kidney agenesis+18178
22tracheo-oesophageal fistula23430
 dysplastic kidney+  
 aortic coarctation  
 distal urethral stenosis and dilatation of the bladder+  
26spina bifida occulta+21400
 malformations genital system, secondary to neural tube defect; aplastic/hypoplastic bladderno bladder visible  
Table 2.  Minor malformations at autopsy.
CaseMinor malformationsDetected with MRIGestational age (weeks)Birthweight (g)
  1. VSD = ventricle septum defect; ASD = artium septum defect.

4VSD 1–2 mm291225
12horseshoe shaped kidney29695
 umbilical hernia+  
18tumour under left kidney (1.5 cm) (neural tissue)18178
20lung lesions+393400
21VSD 1–2 mm20350
25small ASD22738

Two cases of congenital pneumonia and one case of intrauterine sepsis were seen at autopsy. On MRI, one of the two cases of congenital pneumonia showed multiple lung lesions that were interpreted as a substrate for the lung infection. (The other case showed no specific findings on MRI.) In 14 cases, no malformations or only minor malformations were detected at autopsy. In only one of these cases, a malformation was suspected on MRI (abdominal wall defect) that was not established at autopsy. Considering only the major malformations, the diagnostic accuracy of MRI was 17/25. Accepting autopsy as gold standard, the positive predictive value of MRI for the detection of malformations per case of perinatal death was 80% and the negative predictive value was 65% (see Table 3).

Table 3.  Full agreement between autopsy and MRI.
  1. §Cases in which no major malformations were found at autopsy.

  2. *All major malformations in one case detected with MRI.

MRI −71320

The autopsy rate in the study group was 45% (n= 26). MRI was consented to by 59% of all parents (n= 34). There were no cases with consent to autopsy only without consent to MRI. Four pregnancies were electively terminated for congenital malformations. In 39% of cases, the MRI was performed within 24 hours after death or delivery (in case of a stillbirth); in 45%, it was performed between 24 and 48 hours. All scans were performed within 96 hours. In two cases, no consent to MRI was given because the projected time interval between delivery and the scan was considered too long by the parents. In these cases, the scan could have been performed within 48 hours.

A higher acceptance rate for both examinations was seen in the early gestational age group. Before 24 weeks, the parents were more likely to consent to both examinations. In the combined consent group, more fetuses died before delivery but this difference was not significant (n= 20 versus n= 5).


From a diagnostic point of view, accuracy of postmortem MRI in perinatal death was somewhat disappointing even if attention was restricted to major malformations. Several major malformations were clearly visible on MRI (e.g. the presence of dysplastic kidneys in a small fetus) (Fig. 2). In addition, a hydrops fetalis could be well detected (Fig. 3). However, several differences were found between the outcome of autopsy and MRI. Some of these differences can be explained by the early gestational age or low birthweight of the child (see Table 1), some by the nature of the malformation. Cardiac defects were not detected in this study. The same poor sensitivity for cardiac malformations was found in the studies of Woodward et al.11 and Brookes et al.12. In their studies, good results are described in the detection of central nervous system malformations; in our study, the number of central nervous system malformations was too small to confirm this conclusion. Obstetricians as well as parents should be aware that a normal MRI does not necessarily equate with a normal baby; even major malformations can be missed.

Figure 2.

Dysplastic kidneys (T1-weighted image).

Figure 3.

Hydrops fetalis of unknown origin. This child was born at gestational age of 32 weeks with an estimated time of death two days before delivery; the birthweight was 4000 g (T1-weighted image). Hydrops can be seen clearly in the thorax, in the stomach and under the skin.

Only one ‘false positive’ finding was detected at MRI: a presumed abdominal wall defect. This suspected malformation was present in a fetus of 90 g, born at a gestational age of 23 weeks who died intrauterine two weeks before delivery and who was third degree macerated. There was a strange configuration of the umbilical cord seen at autopsy, which could explain the possible abdominal wall defect suspected at MRI.

In the present study, the radiologist was blinded to the clinical findings and did not see the child himself. At least three malformations were missed at MRI that certainly would have been observed after physical examination of the child (anal atresia, omphalocele, myelomeningocele). In addition, the presumed abdominal wall defect would not have been detected. This would improve the positive predictive value to 78% (14/18) and the negative predictive value to 70% (14/20).

Our study population consisted of only four cases of electively terminated pregnancies for congenital malformations. This may be the reason why our rate of malformations is low compared with other studies. In the present study, a total of 18 major malformations were found, compared with 46 major malformations in the study of Woodward et al.11, where the study group of 26 fetuses included 10 terminations of pregnancy for congenital malformations. Therefore, the prevalence, the positive and negative predictive values found in this study, will more accurately reflect an obstetric population in a general district hospital.

In a 13-month cohort of 58 patients, consent for routine autopsy could be obtained in about half of cases, which is in conformity with national and international autopsy rates3,5,6. Consent for MRI was obtained in an additional 15%, demonstrating a higher acceptance. In 84% of all cases, the MRI could be performed within 48 hours after the perinatal death. A faster access to MRI could only have given a slight improvement. The acceptance rate of 59% for MRI in the study group was not divided equally between the different gestational ages at birth. The younger the gestational age at the time of birth, the more likely the parents were to consent to both MRI and autopsy. Especially above 32 weeks, the parental consent to autopsy was low.

No evidence was found in this study for an additional value of MRI when the consent for a full autopsy is obtained. Autopsy remains the gold standard in postmortem examination of perinatal death. Therefore, parents should not be offered MRI scans until it is clear they will not consent to autopsy, except possibly in those cases where an abnormality known to be demonstrable on MRI is suspected.

When offering MRI to parents, they should be adequately counselled on the limitations of the MRI technique, preventing the danger of parental disillusionment if they have the impression that MRI is even or nearly as good as an autopsy. However, the chance of malformations being false positively detected with MRI is so low that, in cases of autopsy refusal, MRI examinations may yield important extra information in relation to the perinatal death.


Financial support for this study was granted by Health Care Research Netherlands. Thanks to Gouke Bonsel for his contribution to this manuscript.