Perinatal post-mortem ultrasound (PMUS): radiological-pathological correlation

Large surveys of parents and healthcare professionals have found that whilst imaging techniques for non-invasive autopsies have a high acceptability [5, 17], one barrier to uptake are concerns about ‘missed diagnoses’, potentially not reaching the same levels of certainty as an ‘invasive’ autopsy [18]. Whilst PMMR has a more established role and its usage is widely published [7, 19‐24], compared to conventional perinatal autopsy findings, the diagnostic accuracy of PMUS is still being evaluated.

Overall PMUS sensitivity and specificity rates have been reported at 75% and 83.3% for whole-body diagnoses [25], with abnormalities affecting the heart being particularly challenging to diagnose with lower sensitivity rates ranging between 18.2–50% [25‐27]. These numbers may be lower than expected or acceptable, although when comparing 2-D PMUS and 3T PMMR imaging within the same cohort of 160 foetuses, Kang et al. [28] did not find a statistically significant difference in accuracy rates for cases where the ultrasound images were of diagnostic quality (67.8% vs 78.0% respectively, p > 0.05). This suggests that a well-performed perinatal post-mortem ultrasound study could provide a similar degree of information as PMMR without the need for further cross-sectional imaging, allowing for potential placement as a first-line technique to triage cases for PMMR. In this article, cases where PMUS was found to be non-diagnostic (i.e. unhelpful) were those that were performed in foetuses of < 20 weeks gestational age [28] and displaying moderate or severe maceration [29]. Therefore, we recommend that operators are aware of the reduced likelihood of a diagnostic PMUS study in small gestational aged foetuses, those that have a known prolonged intra-uterine retention period and those with suspected cardiac anomalies.

Congenital disorders of the central nervous system (CNS) are the commonest group of structural anomalies leading to termination of pregnancy. Approximately 32–37.7% of all terminations of pregnancies may be for prenatally diagnosed CNS anomalies [14, 30], with neural tube defects accounting for the largest proportion (approximately a third of cases) [31].

For this reason, it is important to assess the spine in all foetuses, paying close attention for associated Chiari malformations (Fig. 1) and identifying vertebral segmental anomalies and spinal dysraphism (Fig. 2). Further characterisation with skeletal radiographs may be helpful [32].

The second most common CNS anomaly is lateral ventricular dilatation, or ventriculomegaly. Where prenatal ventricular dilatation is reported, cerebrospinal fluid shifts occurring after death may result in an apparent ‘resolution’ of this appearance [33]. In those instances, PMUS cannot confirm nor refute ventriculomegaly per se but should be used to exclude other contemporaneous anomalies [34] such as callosal anomalies (Figs. 3 and 4), neuronal migration anomalies (Fig. 5) or cerebral aqueductal stenosis, although the latter is harder to identify at post-mortem in the presence of cerebral oedema and sutural distortion.

Congenital intracranial tumours are rare. These account for < 2% of all foetal tumours [35], and teratomas forming the vast majority of all subtypes (Figs. 6 and 7). As a group, congenital intracranial tumours have a poor survival rate (estimated in one study as only 7% in the first year of life [36]), and tumours arising from or extending into the neck are usually at high risk of airway obstruction during delivery (Fig. 8). Intracranial lesions are usually primary tumours rather than metastases, although we have imaged one neonate with multiple intracranial metastases from a congenital fibrosarcoma of the thigh (Fig. 9). Recurrence rates in future pregnancies for intracranial anomalies are fortunately low, particularly where the abnormality is isolated (which is the case for the majority) and an underlying genetic cause is not identified [37].

Congenital cardiac anomalies make up a significant proportion (approximately 10% [31]) of all structural anomalies at termination of pregnancy, and the prevalence of such anomalies are rising. Over the last 20 years across Europe, the annual increase of cardiac anomalies has been estimated at 1.4–4.6% [38] with the reasons thought to be related to increased maternal risk factors such as diabetes, body mass index, assisted reproductive techniques and alcohol consumption. These are key aspects of the mother’s clinical history which should be available at post-mortem imaging.

At PMUS the detection of complex cardiac anomalies is difficult due to a combination of lack of circulating blood, intra-cardiac haemostasis and occasionally intra-cardiac gas (likely from feticide [39]). Some distortion of the normal anatomy at post-mortem examination can be overcome by imaging the foetus in a waterbath [9].

The commonest cardiac anomalies at termination of pregnancy are hypoplastic right/left heart syndrome [31] and uni-ventricular heart defects [40]. Other pathologies also feature, although less commonly, and include pulmonary atresia/stenosis (Fig. 10), aortic valve atresia/stenosis, transposition of the great arteries, tetralogy of Fallot, coarctation of the aorta, anomalous pulmonary venous return and septal defects (Fig. 11).

Where cardiac imaging is non-diagnostic at PMUS, further cross-sectional imaging with PMMR may be useful, particularly if high-resolution, isovolumetric sequences for multiplanar reconstructions are acquired, given the post-mortem distortion of normal anatomy due to ‘slumping’.

Congenital pulmonary anomalies are the least common structural abnormalities seen at PMUS [25]. Whilst congenital pulmonary malformations (including cystic malformations, bronchopulmonary sequestrations, bronchial atresia, congenital lobar emphysema and bronchogenic cysts) may all be seen in live children, these are rarely the cause for foetal demise or terminations of pregnancy [41, 42]. In our experience, we have not detected any airway or lung malformations on PMUS, although Kang et al. [25] report one autopsy confirmed case of a bronchopulmonary foregut malformation in their series, which was missed on PMUS. It could have been due to the subtlety of the appearances that lead to the miss on PMUS; however, the medical literature is sparse with regards to the ideal post-mortem imaging of congenital pulmonary malformations.

The commonest finding at post-mortem imaging of the lungs is lung hypoplasia, usually secondary to other intra-abdominal pathologies such as congenital diaphragmatic hernias (Fig. 12) or enlarged polycystic kidneys. Excluding pulmonary infection is not currently possible [9] given that the foetal and early neonatal lungs are normally fluid filled.

Abnormalities of the abdomen seen at PMUS are most commonly related to the urinary tract or abdominal wall, the latter including pathologies such as gastroschisis, omphalocele (Fig. 13) and congenital diaphragmatic hernia (Fig. 12) [25, 26, 43]. Whilst the presence of an anterior abdominal wall defect does not require ultrasound for diagnosis, the resultant distortion and shift of intra-abdominal organs may have made prenatal imaging difficult and therefore examination of the presence of internal structures is the main criteria for imaging these cases.

Congenital intra-abdominal foetal tumours are very rare but may occur in the liver (such as haemangiomas, mesenchymal hamartoma and hepatoblastomas), kidneys (mesoblastic nephroma), pelvis (sacrococcygeal teratoma) or adrenal gland (neuroblastoma) [44]. We have previously identified splenic metastases from an aggressive primary fibrosarcoma (Fig. 14) and a suprarenal cystic mass secondary to in utero adrenal haemorrhage (Fig. 15).

Renal anomalies are common and account for approximately 6–11% of all structural abnormalities seen at terminations of pregnancy [31, 45, 46]. These are usually related to renal dilatation (Figs. 16, 17, 18, and 19), polycystic renal disease (Fig. 20) or a mixture of structural congenital renal anomalies such as renal agenesis, ectopic kidneys (Fig. 21) or cross-fused ectopia (Fig. 22). A prenatal history of oligohydramnios may be present, and non-visualisation of the urinary bladder at prenatal ultrasound has been reported as a marker for significant renal pathology and poor foetal survival [47, 48].

Where an underlying skeletal disorder is suspected, assessment of the whole skeleton is best performed by external examination and radiography (also known as a skeletal survey or babygram [32, 49, 50]). These are most useful over 8 weeks gestation, given the lack of skeletal ossification prior to this age [51]. It is also important to remember that although isolated limb anomalies may be more commonly encountered in live cases [52], lethal skeletal dysplasias will be more prevalent in the subgroup presenting for post-mortem imaging [53] so an entire overview of the body is usually necessary rather than imaging of the affected limb(s). Whilst PMUS can be useful in demonstrating the non-ossified, cartilaginous anatomy which is not present on radiography (Fig. 23), it is rarely used in isolation to make a skeletal diagnosis and would be better reserved for the assessment of soft tissues.

Soft tissue lesions at PMUS may include venolymphatic (and other vascular) malformations (Fig. 24) or teratomas (as previously shown in Fig. 8) [44, 54]. In cases where a foetus is referred with a prenatal diagnosis of cystic hygroma (usually in the setting of foetal hydrops), we have also found that these cystic masses can ‘resolve’ [55], much like the appearances of ventriculomegaly, thus the role of PMUS is to review associated structural anomalies rather than identify the cystic hygroma itself. Associated anomalies (commonly intracranial in origin) are estimated to occur in 36–55.6% of cases, and usually related to an abnormal genetic karyotype [56, 57] which will have implications for prenatal counselling in future pregnancies [58].