Despite major advances in prenatal diagnosis and neonatal medicine, the ‘gold standard’ for the identification of underlying pathology and the detection of associated abnormalities remains the traditional autopsy in many clinical scenarios. Perinatal autopsy has an established role in confirming and refuting antemortem diagnoses, determining the cause of intrauterine and neonatal deaths and the identification of additional features in abnormal foetuses, which may allow specific clarification of an underlying generic diagnosis. Perinatal autopsy findings are reported to broadly agree with the prenatal clinical diagnosis in 30 to 90% of the cases, depending on the population studied, but change the final diagnosis in 10 to 40%, with additional findings that do not change the clinical diagnosis reported in up to an additional 25% of postmortem examinations. Overall, therefore, in around 30% of the perinatal postmortem examinations, some additional information is obtained from autopsy which changes either the underlying diagnosis or information given to parents during counselling (Gordijn et al., 2002; Amini et al., 2006; Sankar and Phadke, 2006; Akgun et al., 2007; Gordijn et al., 2007). The likelihood of gaining such information is also related to the specialist status of the pathologist, with significantly more findings reported when the autopsy is performed by paediatric specialist, compared to general, histopathologists, and other factors such as the length of time between death and autopsy (Vujanic et al., 1998; Gordijn et al., 2002).
Ongoing advances prenatal diagnosis and increasing use of high-resolution foetal ultrasound (USS) have resulted in improved diagnosis of a range of congenital malformations at earlier gestational ages with increased confidence in their diagnosis. Nevertheless, discrepancy between prenatal USS and postmortem findings still occurs in a significant minority of cases, particularly in relation to specific neurological and cardiac abnormalities, such as ventriculomegaly or posterior fossa defects, and cardiac specific malformations (Gordijn et al., 2002).
However, despite the increasing number of terminations of pregnancies in countries such as England and Wales as a consequence of more widespread and specialised prenatal diagnosis facilities, there has been a reduction in the proportion of parents who consent to a conventional autopsy in recent years, with perinatal autopsy rates being around 50% of eligible cases, well below the recommended minimum standard of 75%, as previously suggested by a joint working party of the Royal Colleges of Obstetricians and Gynaecologists and Pathologists (RCOG, 2001; DoH, 2008). While there are several reasons for this decline in autopsy rates, one major factor is likely to be lack of parental agreement to the procedure based on concerns, primarily regarding disfigurement of the body and organ retention (DoH 2000; Gordijn et al., 2002). In addition, clinicians may not appreciate the continued potential beneficial information which autopsy can provide, may not like discussing postmortem examination, and may be less likely to discuss autopsy if they feel that this could cause parental distress (Epstein, 2008). Even in cases for which parental consent is provided, changes in public perception since the Organ Retention Controversies affecting some hospitals in the 1990s may limit the amount of information available with traditional methods (DoH, 2000; Human Tissue Authority (HTA), 2009). For example, the usual practice for examining the central nervous system is to remove and fix the brain before further dissection, a process that could take 3 to 4 weeks. Parents may now request that all organs are returned to the body before burial without delay, making adequate fixation difficult with associated interpretative difficulties. Adequate neuropathological examination is even more problematic if there is delay between intrauterine death and delivery, leading to maceration of the foetus, and in association with foeticide using potassium chloride; (Carroll et al., 2000) the RCOG guidelines suggest that any pregnancy terminated after 22 weeks of gestation should be accompanied by foeticide to ensure that the foetus is not born alive (RCOG, 1996). In addition to the practical issues of obtaining consent to perform conventional autopsy, and sufficient time to perform optimal histological examination, there is increasing pressure from some religious communities, for whom conventional autopsy may be less acceptable, unless it is a legal requirement, to provide a less invasive, accurate and widely available method of postmortem assessment (Bisset et al., 2002). Public awareness of organs being retained following autopsy in the UK (DoH, 2000), has led to the introduction of new legislation regarding all aspects of the examination and use of human tissue, the Human Tissue Act (HTA, 2009), and, revision of Her Majesty's Coronial system (Coroners Society, 2009), which investigates all deaths of unknown cause and a range of other deaths including some perinatal cases. This legislation has addressed issues surrounding removal and retention of tissue and highlighted the need to critically examine both the role of postmortem examination and how it is carried out.
RATIONALE FOR POSTMORTEM MAGNETIC RESONANCE IMAGING AND CT SCANNING
Despite the potential benefits noted earlier, for routine clinical practice in many countries, offering an optimal perinatal postmortem service may not be easy or practical due to a lack of appropriately trained pathologists in specialist centres. (RCPCH, 2002) Furthermore, these services may involve both transfer of the body to a geographically distant specialist centre and increased delay due to the need for specialist referral, with the findings being complicated by superimposed effects of postmortem autolysis and decomposition. As a significant component of the traditional autopsy involves macroscopic examination and dissection of internal anatomy, use of a noninvasive radiological technique that can provide good soft-tissue imaging, such as a magnetic resonance imaging (MRI), is a potential candidate to either augment the traditional autopsy or modify its approach. There are several potential advantages to postmortem imaging compared to the performance of MRI in life, which allow enhanced tissue visualisation in this particular setting. First, there is no movement of the patient or region of interest (ROI) and therefore no motion artefacts are present, and, since there are no limits on acquisition time, very good spatial resolution and image quality are possible. Secondly, as the ROI in perinatal cases is relatively small, the receiver coil can be optimally placed to allow maximum definition and resolution to be achieved. Furthermore, the ROI can be precisely placed in the optimum imaging position. Using appropriately modified acquisition protocols to optimise visualisation in the postmortem setting, excellent tissue detail can be obtained from many structures (Figure 1).
A further, but highly significant, advantage of the use of cross-sectional imaging modalities is the acquisition of a large raw data set, which can subsequently be manipulated and reanalysed if required and opinions obtained from other specialists if necessary.
Despite the good soft-tissue visualization of MRI, bony imaging is relatively poor, whereas computerised tomography (CT) provides rapid and excellent imaging of bony skeletal structures and also allows similar acquisition of data for further three-dimensional (3D) reconstruction as required (Figure 2). The combination of postmortem MRI and CT scanning is likely to provide maximal diagnostic data, with the development of additional sequences or techniques, such as diffusion weighted MRI or magnetic resonance (MR) spectroscopy, being added as experience increases to address specific issues.
STUDIES REPORTING ON FOETAL PM MRI
It is now more than a decade since the use of whole body PM MRI in foetuses was reported as an alternative to conventional autopsy for parents who refused a traditional postmortem examination (Brookes et al., 1996). This proved to be a controversial publication and several groups responded to highlight the shortcomings of an imaging-based approach suggesting that postmortem examination should only be performed by the traditional method (Berry et al., 1997; Niermeijer, 1997; Rushton, 1997).
Nevertheless, several research groups have subsequently performed PM MRI scanning in groups of selected cases and have reported its use in small studies (Woodward et al., 1997; Huisman et al., 2002; Alderliesten et al., 2003; Griffiths et al., 2003; Breeze et al., 2006; Brookes and Hagmann, 2006; Widjaja et al., 2006; Cohen and Whitby, 2007; Hagmann et al., 2007; Cohen et al., 2008; Table 1). With recent renewed interest in the possibility of the development of less invasive postmortem examination, for the reasons described earlier, the need for evidence regarding the accuracy of PM MRI has regained its importance. Published data to currently support the use of PM MRI are primarily based on relatively small studies with varied inclusion criteria (Table 1). Several reports include data sets presented in previous publications, and the extent of blinding of radiologists and pathologists to the MRI and autopsy results are uncertain (Table 1). For accurate assessment of diagnostic use of a test, it is essential that the index test (PM MRI) and reference test (traditional autopsy) are reported independently of each other. The issues with these published data have arisen primarily because most studies have been performed retrospectively, partly due to the practical difficulties of performing prospective research in these circumstances, (Thayyil et al., 2008) and partly as the usual consequence of the development of most new techniques into clinical practice, which starts as the technique evolves with selected cases. Moreover, the interpretation of the findings is made more difficult as it may not be easy to clearly differentiate between postmortem ‘artefacts’, which have no clinical significance, and true clinically important imaging/pathological findings. Finally, the choice of specific and appropriate MRI sequences, optimised for the postmortem setting, is not clearly provided, all previous studies have used low/normal field strengths (1.5 T), and the expertise for reporting of 3D MR images in foetuses may be variable. Currently, therefore, while PM MRI appears a promising technique, its cost-effectiveness and accuracy in unselected populations remain uncertain.
Table 1. Summary of the main studies reporting on postmortem MR imaging in foetuses
10 cases with a spinal abnormality noted on prenatal USS and 31 foetuses without abnormality included. Postmortem MR detected abnormality in all cases. Selected cases and blinding of radiologists and pathologist unclear
Retrospective review with inclusion of previous cases. 60% agreement between MR and autopsy findings
Additionally, it is not always clear whether published studies have assessed the role of PM MRI as a useful alternative or adjunctive test to conventional autopsy. As these are very different purposes they would require different study designs for their appropriate evaluation. For example, to be an accurate alternative investigation, the diagnostic accuracy of PM MRI must be at least similar to conventional autopsy with regard to major findings, whereas to be a useful adjunctive investigation, the PM MRI should simply detect any findings that may be clinically significant, but which may be missed or difficult to detect on conventional autopsy, or could instigate further investigations. Although no study has specifically addressed this issue, it has been clearly suggested (Brookes et al., 1996); a significant proportion of parents may not agree with traditional autopsy examination and in this group, the PM MRI represents neither an alternative nor an adjunct to traditional autopsy, but an additional approach that may allow at least some important clinical information to be obtained. In this population, the alternative is no form of examination at all and therefore, to be regarded as ‘useful’, the PM MRI simply has to provide some additional diagnostic information compared to no examination whatsoever.
Recently, in the UK, the Chief Medical Officer ordered a review of the role of less invasive autopsy and the Department of Health has funded two separate large comparative trials of PM MRI in relation to conventional autopsy for both adults and foetuses/infants. These are planned for completion in 2010. The foetal/infant arm of this study (MaRIAS: Magnetic Resonance Autopsy Imaging Study) is being performed at centres in London, representing a collaborative group of radiologists, foetal medicine specialists, geneticists, paediatric pathologists, neonatologists, MR physicists and statisticians. Each MR image is examined and reported in a structured way by specialist paediatric radiologists for all organ systems, following image acquisition using specially optimised sequences specific to foetal PM imaging. Based on existing experience in living patients, it is likely that MRI will allow accurate detection of many structural abnormalities, and its sensitive determination of internal organ volumes has already been reported (Thayyil et al., 2009a). However, analysis of the MaRIAS trial data to determine the true sensitivity and specificity of PM MRI compared to traditional autopsy in this paediatric setting is not yet available.
CURRENT ISSUES WITH FOETAL PM MRI
Difficulties imaging skeletal structures: PM CT
CT imaging is cheaper and significantly faster than MR imaging. However, image contrast in postmortem CT for visceral organs is relatively poor compared to MRI, thus, the use of CT alone as an alternative for anatomical examination in foetal and infant autopsies is also poor. Nevertheless, for foetuses with skeletal malformations and other bony abnormalities, CT imaging and 3D reconstruction may be superior to conventional X-ray examination, as accurate data sets allow full 3D reconstructions with excellent visualisation of the entire skeletal system (Figure 2).
Difficulties imaging small foetuses: high-field MRI
Standard PM MRI has been performed using widely available diagnostic MR scanners with 1.5 T magnetic field strengths. This methodology allows good image acquisition in most foetuses beyond 20 weeks of gestation, but due to technical issues regarding the size of the ROI results in suboptimal resolution of anatomical detail in foetuses in the earlier second trimester (Thayyil et al., 2009b). MRI at high-field strengths (9.4 T) remains at present a research technique but has opened an avenue of ‘virtual microscopy’ in animal models, with resolutions of up to 20 µm being achieved in some cases (Fatterpekar et al., 2002). Whole body PM MRI of foetuses at 9.4 T is possible and results in outstanding resolution of detail even in small foetuses. In a recent small pilot study of the technique, structural defects identified at autopsy could be detected by high-field MR imaging (9.4 T), whereas conventional (1.5 T) MR imaging was non-diagnostic in most of the cases of foetuses delivered at < 20 weeks of gestation. Good images could be obtained even in foetuses as small as 5 g. High-field MR imaging may therefore in future allow adequate structural examination of even very small foetuses and may allow highly specific diagnosis of organ changes in full term foetuses and newborn infants (Thayyil et al., 2009b; Figure 3). The main current limitation of this technique is that such high-field scanners are currently available only in few tertiary research centres in the UK and require special expertise. Nevertheless, with increasing development of specialist perinatal autopsy networks in selected centres in the UK, transfer of bodies could be organised should the use of high-field scanners become more widespread.
The major disadvantage of any imaging-based ‘minimally invasive’ autopsy approach is the lack of availability of tissue samples for subsequent microscopic histopathological examination or other ancillary techniques, such as microbiological, genetic or molecular studies, which often provide the most important and useful clinical information. The adjunctive use of postmortem image-guided needle tissue biopsies may allow the collection of such tissue samples, because needle-core biopsy is being increasingly used for the primary investigation of a large range of paediatric conditions in life with sufficient diagnostic material obtained (Sebire and Roebuck, 2006). However, the adequacy of such sampling in the postmortem setting has not yet been established, in particular regarding the potential issues of sample representativeness. In one study of 30 cases, PM needle biopsies were performed of major organs following PM MRI but not under direct MR image guidance, but representative and sufficient material was obtained in only 15 (50%; Breeze et al., 2008). A more practical effective alternative approach would be to perform PM imaging followed by endoscopic/laparoscopic direct organ visualization with biopsies carried out under direct vision, allowing macroscopic organ appearances to be determined in addition to providing larger biopsy samples. The feasibility and acceptability of this approach remain undetermined at present.
Development of the specialty of PM imaging: who should report the findings?
It is inappropriate to compare postmortem MR imaging diagnoses alone with those based on the full range of traditional postmortem investigations, which include clinical history, external examination for dysmorphic features, placental histopathology, etc. It should be recognised that PM imaging, by any method, can provide information primarily on structural abnormalities and it is therefore important to emphasise that the concept of the ‘minimally invasive autopsy’ represents a collection of approaches that involve all noninvasive parts of postmortem examination, including external examination of the body, photography, imaging, placental and foetal/infant histopathological examination and genetic testing, but without the traditional open dissection for internal examination of the visceral organs. In this context, the appropriate reporting of postmortem MR imaging should be incorporated into a cohesive unified autopsy report, rather than an imaging report in isolation. Consistent with introduction of all new medical techniques, as a consequence of the issues of both optimising sequences for data acquisition and interpretation of findings in the postmortem setting compared to during life, there will be an initially steep learning curve for those involved in the development of this service. This is particularly so since a wide range of ‘normal’ postmortem changes seen at autopsy as a consequence of changes following death can mimic diseases in life and such changes must be recognised and accounted for to allow correct interpretation. Appropriate reporting in this context will therefore require large-scale validation of imaging techniques against the currently accepted gold standard of the traditional autopsy, but at present it remains uncertain regarding which medical professionals should take the lead in this role. The foetal medicine specialist has expertise in the diagnosis of foetal anomalies and the interpretation of two-dimensional and 3D ultrasound data sets, but usually little experience or training in understanding the process of death and associated postmortem changes, or recognition of organ or disease-specific pathologies. Paediatric pathologists are skilled in the overall examination of foetuses and infants, in particular, the correct interpretation of additional investigations and histopathological changes, but most have no training or expertise in the interpretation of diagnostic imaging modalities. Radiologists with an interest in MRI are clearly best equipped to optimise the technical aspects of image acquisition and manipulation, but almost none are trained in aspects of foetal medicine, pathologies that may affect the foetus and infant, or the pathological aspects surrounding death. It is clear that the introduction of less invasive postmortem examination in any format will require a multidisciplinary team approach and, due to the requirement to coordinate and collate the results of all investigations, it is likely that, paediatric pathologists, with input from specialist radiologists with expertise in postmortem imaging, together with other clinicians such as geneticists and foetal medicine specialists, will continue to act as the main specialist at present.
Postmortem imaging with MRI/CT represents a new development towards the concept of minimally invasive autopsies and may allow diagnosis of abnormalities in cases where parents would not agree with a traditional open autopsy, in addition to providing additional diagnostic information when used as an adjunct to the traditional autopsy. At present however, until adequate evidence on the accuracy and cost-effectiveness of such an approach is available, the use of postmortem MR imaging as a replacement for conventional autopsy cannot be recommended. However, it is almost certain that the use of PM imaging will rapidly become an important integral component of postmortem examination of the foetus and infant in specialist centres, which may determine the type and extent of further examination required for a given case. In addition, ‘minimally invasive’ autopsy examination could provide clinically relevant additional information for those parents who decline traditional autopsy, and hence at present do not undergo any type of postmortem foetal examination. The concept of the minimally invasive autopsy is likely to gain more widespread acceptance with changes in public and parental perceptions of the autopsy.