Diagnostic quality of 3Tesla postmortem magnetic resonance imaging in fetuses with and without congenital heart disease

BACKGROUND: Postmortem conﬁrmation of prenatally diagnosed congenital heart disease after termination of pregnancy and evaluation of potential cardiac defects after spontaneous fetal or neonatal death are essential. Conventional autopsy rates are decreasing, and 1.5Tesla magnetic resonance imaging has demonstrated limited diagnostic accuracy for postmortem cardiovascular assessment. OBJECTIVE: This study aimed to evaluate the feasibility and image quality of cardiac 3Tesla postmortem magnetic resonance imaging and to assess its diagnostic accuracy in detecting fetal heart defects compared with conventional autopsy. Secondarily, the study aimed to explore whether clinical factors affect the quality of 3Tesla postmortem magnetic resonance imaging. STUDY DESIGN: A total of 222 consecutive fetuses between 12 and 41 weeks’ gestation, who underwent 3Tesla postmortem magnetic resonance imaging and conventional autopsy after spontaneous death or termination of pregnancy for fetal malformations, were included. First, 3Tesla postmortem magnetic resonance imaging of each fetus was rated as diagnostic or nondiagnostic for fetal cardiac assessment by 2 inde-pendent raters. The image quality of individual cardiac structures was then further evaluated by visual grading analysis. Finally, the presence or absence of a congenital heart defect was assessed by 2 radiologists and compared with autopsy results. RESULTS: Overall, 87.8% of 3Tesla postmortem magnetic resonance imaging examinations were rated as diagnostic for the fetal heart. Diagnostic imaging rates of individual cardiac structures at 3Tesla postmortem magnetic resonance imaging ranged from 85.1% (atrioventricular valves) to 94.6% (pericardium), with an interrater agreement of 0.82 (0.78 e 0.86). Diagnostic imaging of the fetal aortic arch and the systemic veins at 3Tesla postmortem magnetic resonance imaging was possible from 12 þ 5 weeks’ gestation onward in 90.1% and 92.3% of cases, respectively. A total of 55 fetuses (24.8%) had at least 1 cardiac anomaly according to autopsy, 164 (73.9%) had a normal heart, and in 3 fetuses (1.4%), autopsy was nondiagnostic for the heart. Considering all examinations rated as diagnostic, 3Tesla postmortem magnetic resonance imaging provided high diagnostic accuracy for the detection of fetal congenital heart defects with a sensitivity of 87.8%, a speciﬁcity of 97.9%, and concordance with autopsy of 95.3%. 3Tesla postmortem magnetic resonance imaging was less accurate in young fetuses ( < 20 weeks compared with (cid:2) 20 weeks; P < .001), in fetuses with low birthweight ( (cid:3) 100 g compared with > 100 g; P < .001), in cases after spontaneous fetal death (compared with other modes of death; P ¼ .012), in cases with increasing latency between death and 3Tesla postmortem magnetic resonance imaging ( P < .001), and in cases in which there was a high degree of maceration (maceration score of 3 compared with a score from 0 to 2; P ¼ .004). its high diagnostic accuracy, suggest 3Tesla postmortem magnetic imaging may serve as a suitable with which to direct a targeted conventional when are or to a virtual when full is the parents.


Introduction
Congenital heart disease (CHD) is the most common malformation in fetuses and newborns, with a prevalence of 8 per 1000, comprising live births, fetal deaths after 20 weeks' gestation, and terminations of pregnancy for fetal malformations. 1 In Europe, up to 44% of pregnancies are terminated prematurely owing to prenatally diagnosed CHD, either in isolation or as part of complex syndromes. 1,2 Indeed, the diagnostic quality of prenatal screening for CHD varies widely owing to the limited availability of specialists. 2,3 Moreover, the diagnostic accuracy of cardiac ultrasound (US) is limited per se, even when performed by specialists. 2,4,5 The prenatal diagnosis of CHD has been improved recently by using intelligent methods, such as fetal intelligent navigation echocardiography, with more than 90% diagnostic accuracy, but these techniques are not yet widely used. 6 Finally, only a relatively small proportion of fetuses undergo targeted cardiac US for suspicion of CHD. Accordingly, the cause of spontaneous death in utero is often unclear.
A postnatal, full workup of fetal malformations is essential: even if prenatal US suggests CHD, this suspected diagnosis warrants postnatal confirmation or clarification. In case of fetal loss, stillbirth, and neonatal death related to nonchromosomal fetal anomalies, autopsy adds important clinical information about etiology and individual recurrence risks in approximately 25%. 7 Because cardiac malformations as a potential expression of a hereditary disease can affect further pregnancies, these must be classified as high-risk pregnancies and require special care.
Autopsy remains the gold standard of postmortem evaluation to confirm and to complement prenatal imaging results and to detect missed cases of CHD. 4,8 However, both clinicians and parents increasingly less often ask for and consent to autopsy. 9e11 The reasons for this development are diverse and include a physician's limited ability to select cases for which autopsy will have the greatest diagnostic yield, and the parents' lack of information about the possible benefit from autopsy. Another important reason is the increasing workload of pathologists, together with vanishing expertise in the field of perinatal pathology outside of tertiary centers. 11,12 Accordingly, virtual and minimally invasive autopsies have emerged as complementary investigations to conventional autopsy owing to advancements in imaging technology and rapidly increasing demand. 9,10,13,14 Postmortem magnetic resonance imaging (pmMRI) has become a relevant virtual addition to conventional autopsy or is even considered an alternative, because it is less invasive and more readily accepted by the parents. 9,14,15 Previous small studies have evaluated the diagnostic value of pmMRI for fetal cardiac evaluation. 9,15e19 The diagnostic accuracy of 1.5 Tesla (T) pmMRI in small fetuses at less than 20 to 24 weeks' gestation is lower than 50% with regard to the fetal heart. 9, 17,20 The better imaging quality of 3T cardiac pmMRI has been reported, 14e17 with fewer nondiagnostic examinations (11.6% in 3T vs 32.7% in 1.5T), higher sensitivity (48% vs 36%), specificity (95.1% vs 70.5%), and concordance rates (81.4% vs 60.5%) in the detection of abnormalities for 3T pmMRI than 1.5T pmMRI, particularly in fetuses younger than 20 weeks' gestation. 17 This study comprehensively examines cardiac 3T pmMRI.
This study sought to determine the feasibility and the diagnostic value of 3T pmMRI of the heart and great vessels in a consecutive series of fetuses with and without CHD compared with autopsy results. Furthermore, the influence of clinical factors, such as birthweight and the time span between fetal death and 3T pmMRI, on MRI image quality and diagnostic accuracy was assessed.

Study participants
This retrospective, single-center study was conducted at the Medical University Hospital of Vienna, Austria, at the Departments of Obstetrics and Fetomaternal Medicine, Biomedical Imaging and Image-Guided Therapy, and Pathology. The study protocol was approved by the institutional ethics committee (EK1658/ 2018). A full, informed, written consent was obtained from all patients before 3T pmMRI and autopsy. We evaluated all consecutive 3T pmMRI examinations performed between 2012 and 2019 of fetuses who had died pre-or perinatally at the Department of Obstetrics and Fetomaternal Medicine. All fetuses with a prenatal US evaluation of the heart, 3T pmMRI, and conventional autopsy were included. Included fetuses and neonates died spontaneously before, during, or immediately after birth or owing to termination of pregnancy (TOP) for structural or genetic abnormalities. Only cases with pharmaceutical induction of labor or cesarean delivery were included. For TOP, mifepristone followed by misoprostol was used to induce labor. Intracardiac injection of lidocaine 1% was used for feticide after week 23, before induction of labor with prostaglandins. Detailed characteristics of the study population are provided in Supplemental Table 1.

Postmortem magnetic resonance imaging and image analyses
Fetuses were kept in a refrigerator at 4 C from birth until 3T pmMRI. 3T pmMRI examinations were performed outside of routine office hours in order not to disturb clinical routines, within 24 hours after delivery. Acquisition of the 3T pmMRI examination depended mainly on the availability of radiologists and MRI scanners in the evening and during the night; in any case, the corpse was brought to the Department of Pathology for full autopsy after 3T pmMRI within 24 to 48 hours.
All pmMRIs were performed on a 3T scanner (Magnetom Trio and Magnetom AJOG at a Glance Why was this study conducted? This study aimed to evaluate the feasibility, image quality, and accuracy of cardiac 3Tesla postmortem magnetic resonance imaging (3T pmMRI) in a consecutive series of fetuses for the detection of congenital heart defects compared with autopsy and to explore whether clinical factors affect the quality of postmortem imaging.
Vida, Siemens Medical Systems, Erlangen, Germany) using a knee coil, flex coil, or body coil according to the size of the fetus. To date, examination protocols for the body, excluding the brain, have consisted of T2-weighted images (with the field of view [FOV] depending on the region) of the body in at least 2 orthogonal planes (axial and coronal). The sagittal plane and oblique planes were reconstructed from the original data set (3- To assess the feasibility and the image quality of fetal cardiac pmMRI and to detect potential cardiac malformations, 2 experienced radiologists (4 years and 11 years of experience in cardiac imaging; 2 years of limited experience in postmortem imaging), who were blinded to any clinical information (US findings, gestational age, mode of death, autopsy results), independently rated different anatomic structures of the fetal heart and the great vessels on a picture archiving and communication system (AGFA).The following structures were evaluated: left and right atria, left and right ventricles, atrial and ventricular septum, atrioventricular valves, left and right heart outflow tracts, aortic origin and aortic arch, pulmonary arteries including the ductus arteriosus Botalli, pericardium, pulmonary veins, and venae cavae.
Each cardiac structure was scored independently, receiving a score of 1 if the anatomic structure was clearly visible and assessable without restriction (regardless of normal or abnormal); a score of 2 if visibility was limited, but assessable with restriction; and a score of 3 if the structure was not clearly visible and not assessable (nondiagnostic). 21 Interrater agreement and Cohen's kappa values were calculated from the individual scoring results of the 2 independent raters (step 1). Cardiac structures were grouped together when the specific ratings of 1 structure (such as the mitral valve) were always identical with the ratings of the anatomically similar or corresponding structure (such as the tricuspid valve). Only cardiac structures that were visible simultaneously, provided that the correct axes could be displayed, were grouped together. In cases where the 2 raters scored the visibility of cardiac anatomic structures as diagnostic, but at different levels, consensus was then obtained in a second scoring round and a final score was calculated (step 2). Nondiagnostic images (score 3 on our scale) were noted when 1 or both raters considered that the respective cardiac structures were not assessable enough to be rated as diagnostic (no consensus required). When a CHD was noted, the specific suggested cardiac malformation diagnoses, based on 3T pmMRI, were noted in agreement between the 2 radiologists (1 with less and 1 with more experience in cardiac imaging; step 3).

Pathology
A senior fetal pathologist performed the autopsy in all cases according to a predesigned protocol (Supplementary Material: Protocol for fetal and perinatal autopsies). Only when all personnel involved in patient care agreed that there was no scientific or legal reason for autopsy and no relevant information regarding the cause and circumstances of death was to be expected from invasive postmortem examination, the parents were given the choice to forego autopsy. Autopsy reports were reviewed by a pathologist blinded to US and pmMRI records, but aware of the maternal obstetrical history, gestational age, and mode of fetal death.
The primary endpoints of analyses were the feasibility and the image quality of 3T pmMRI for the depiction of individual fetal cardiac structures. Feasibility was based on binary discrimination of diagnostic vs nondiagnostic examinations of individual cardiac structures at 3T pmMRI. Image quality was assessed using a 3-stage visual grading scale and rated very good to good, moderate but diagnostic, and nondiagnostic imaging of individual cardiac structures at 3T pmMRI.
Secondary aims were to assess diagnostic accuracy of 3T pmMRI in the detection of fetal CHD compared with autopsy and, furthermore, to evaluate the clinical factors that could affect the rate of diagnostic examinations at 3T pmMRI and concordance with autopsy results. Concordance with autopsy results was recorded when both 3T pmMRI and autopsy agreed about the presence or absence of a fetal or neonatal CHD, without further consideration of whether the specific CHD had been diagnosed correctly at 3T pmMRI. Clinical factors assessed for their influence on feasibility and concordance rates were gestational age, birthweight, the reason for fetal or neonatal death (if known), and the circumstances of death (pharmaceutical TOP with and without feticide or spontaneous death; live birth vs stillbirth), the presence or absence of extracardiac malformations, and time between death and 3T pmMRI.
We intentionally did not compare the diagnostic quality of 3T pmMRI vs prenatal US, because this was not the primary aim of our study, and fetal echocardiography reports were inconsistent and oftentimes from substantially earlier gestational ages than fetal death, 3T pmMRI, and autopsy.

Results Demographics
Our initial data set consisted of 262 consecutive fetal 3T pmMRIs from 2012 to 2019. A total of 40 fetuses (40 of 262, 15.3%) were excluded owing to missing prenatal US reports (14), missing wholebody sequences in 3T pmMRI (15), or unavailable autopsy reports (11). The data from 222 fetuses were included, of which 55 (55 of 222, 24.8%) had 1 or several structural heart defects and 164 (164 of 222, 73.9%) had a normal heart according to autopsy. The characteristics of fetuses included are summarized in Table 1.
Concordance rates between 3T pmMRI and conventional autopsy were significantly lower after spontaneous intrauterine fetal death than other modes of death (P¼.012). Stillbirth was associated with lower concordance rates than live birth (136 of 170 vs 47 of 52; 80.0% vs 90.4%; P¼.085). Increasing time intervals between death and 3T pmMRI decreased concordance rates progressively (P<.001). Third-degree maceration was associated with lower concordance rates than maceration scores of 0 to 2 (P¼.004).
Diagnostic accuracy was based on the data from 219 fetuses in which conventional autopsy reports were diagnostic. Nondiagnostic ratings from cardiac 3T pmMRI were considered false negative. We also assessed the diagnostic accuracy (n¼192; pmMRI) when nondiagnostic ratings were excluded from calculation. Diagnostic accuracy data for both The spectrum of cardiac malformations diagnosed on 3T pmMRI and prenatal US are presented in Table 4. Normal fetal cardiac images in 3T pmMRI are displayed in Figures 3 and 4, and postmortem CHDs in fetuses from weeks 14 to 23 are shown in Figure 5. Details of cardiac pathologies are given in Supplemental Table 2, and the reasons for nondiagnostic ratings in 3T pmMRI are summarized in Supplemental Table 3.

Principal findings
3T pmMRI is a feasible imaging modality with which to assess the fetal heart and great vessels postmortem. Furthermore, 3T pmMRI provides high diagnostic accuracy for a wide spectrum of cardiac anomalies compared with conventional autopsy. The rate of nondiagnostic images and, subsequently, impaired diagnostic accuracy in the detection of CHD is related to factors, such as low birthweight (below 100 g), extracardiac thoracic anomalies, and advanced maceration. 16,17,19

Results
We were able to validate the feasibility of 3T pmMRI of the fetal heart and great vessels regardless of gestational age, in the largest series of 222 consecutive fetuses thus far. Depiction of individual cardiac structures was feasible in more than 9 of 10 fetuses from 16 gestational  Values are expressed as number (percentage), interrater agreements (percentage) and Cohen's kappa (95% confidence intervals). Image quality was rated independently by 2 radiologists as 1 for very good to good visibility of the structure(s), 2 for moderate but diagnostic, and 3 for nondiagnostic examination. Diagnostic examinations at 3T pmMRI are expressed as the sum of diagnostic scorings (scores 1 and 2), when none of the 2 raters stated the assessment of the respective cardiac structures as nondiagnostic (score 3). Interrater agreements and Cohen's kappa values are based on independent scorings.  However, image quality and, therefore, diagnostic accuracy depend on various factors. A deformed body shape after having wrapped the fetus for conservation hampers the differentiation of individual organs. Other limitations are harder to overcome, such as the lack of blood in the ventricles postmortem, which interferes with diagnostic imaging of the complete 4-chamber view. Intrathoracic abnormalities, such as diaphragmatic hernia or pleural effusion, have a similar detrimental effect on the feasibility of cardiac 3T pmMRI, because the side walls are squeezed toward the septa, which hampers the distinct imaging of both septa and atrioventricular valves. Furthermore, a well-designed protocol for 3T pmMRI, with planes adjusted to the cardiac axes and the outflow tracts, is essential, particularly with regard to their small size. Evaluation of the intrathoracic aorta is particularly difficult when the arch is dislocated or when a large arterial duct mimics the features of isthmus stenosis,  Diagnostic quality rates of cardiac postmortem MRI for the detection of fetal congenital heart defects Open white circles represent diagnostic quality rates when nondiagnostic images were rated as false-negative and blue circles represent the respective values when nondiagnostic images were excluded.

Clinical implications
The diagnostic accuracy of 3T pmMRI depends substantially on the decision about how to proceed with nondiagnostic examinations. Long latencies from demise to 3T pmMRI and autopsy, such as in cotwin demises with continuation of the pregnancy, are associated with high nondiagnostic 3T pmMRI rates. Removing those cases would raise the performance of 3T pmMRI in cases with a >3-day latency to be similar to those cases with a 3-day latency (70% vs 67%). When nondiagnostic cardiac 3T pmMRI examinations were considered false negative, the overall concordance of 3T pmMRI with autopsy in diagnosing a fetal CHD was 83.6%. Exclusion of nondiagnostic 3T pmMRI imaging led to significantly higher diagnostic quality rates, with a sensitivity of 87.8%, a specificity of 97.9%, and a concordance of 95.3%. We intentionally included all consecutive fetuses with 3T pmMRI and autopsy reports available to investigate clinical feasibility and diagnostic accuracy in the detection of CHD in the daily routine. Most previous studies either excluded nondiagnostic pmMRI imaging or included only fetuses without additional malformations or highly macerated bodies. 15,18,19 In our opinion, this does not reflect the clinical situation of diverse fetal disease. Similar to other authors, we found the diagnostic value of cardiac 3T pmMRI less accurate in younger fetuses of lower birthweight. 16e19 Furthermore, we observed high rates of nondiagnostic cardiac 3T pmMRI examinations in fetuses after spontaneous intrauterine death with a high degree of maceration, which resulted in high discordance rates (nondiagnostic 3T pmMRI vs diagnostic autopsy) in these fetuses. For fetuses younger than 20 weeks, 9.4T pmMRI 18 and microfocus computed tomography (CT) 21e24 have been suggested. Indeed, micro-CT scanners and high-field MRI scanners are not widely available. The procedure of minimally invasive has about the same diagnostic accuracy for the detection of major abnormalities as conventional autopsy, even in younger fetuses. However, undetected abnormalities were predominantly located in the lungs and the heart, most frequently reported in young fetuses at 24 weeks' gestation or less. 9 The approach with 3T pmMRI provides sufficient image quality with high resolution of anatomic structures, is more widely accessible, and allows a suitable postmortem examination both for young and for older fetuses with the advantage of a high soft-tissue resolution compared with CT.

Research implications
Fetal postmortem evaluation is crucial, and considering the decline in parental consent for invasive investigation and the increased pathologist's workload, targeted autopsy in selected fetuses may aid both families and clinicians. 3T pmMRI provides almost the same degree of accuracy as conventional autopsy and is widely available, and images may be transferred from anywhere to specialized centers with expertise in pre-and perinatal medicine. Considering these facts, we urgently need further studies in this field to upgrade postmortem imaging as an integral part of the clinical routine.

Strengths and limitations
Previous investigations of the diagnostic accuracy of pmMRI excluded a substantial proportion of fetuses owing to unavailable autopsy records. 14,25 A major strength of this study is the inclusion of a large, consecutive cohort of fetuses with 3T pmMRI, of which only 4.2% had to be excluded owing to missing autopsy data. In Austria, the law permits nonforensic autopsies without family consent when there is a medical or scientific interest; therefore, autopsy rates in Austria are high compared with other countries. 26 The major limiting factor for cardiac evaluation in this study was the inhomogeneous 3T pmMRI examination protocols adjusted to extrathoracic body regions, predominantly the fetal brain. Only 51.8 % of fetuses

Conclusions
3T pmMRI is a feasible tool with which to evaluate the fetal heart and to detect CHD at any gestational age. Diagnostic imaging rates were particularly high in fetuses with a birthweight of >100 g, a gestational age of >20 weeks, low-degree maceration, and after a <3-day interval from death to pmMRI. After excluding nondiagnostic imaging, specificity for the detection of CHD was high, especially with regard to nonseptal cardiac anomalies. Cardiac 3T pmMRI may act as a decision tool as to whether or not to proceed with invasive postmortem investigation, such as conventional autopsy.   Imaging parameter for cardiac 3Tesla postmortem magnetic resonance imaging on the 3T scanner SIEMENS MAGNETOM VIDA-XT-128, Siemens, Erlangen, Germany. Examination protocols were adjusted to the pathologies found in prenatal ultrasound and previous fetal magnetic resonance imaging examinations. In the first 18 months of postmortem imaging we used a T2weighted sequence in the axial and coronal plane to assess the body. Since the end of 2013, we in addition used a 3dimensional (3D) constructive interference in steady state (CISS) sequence enabling multiplanar reconstructions in various planes.
C. Further cardiac pathologies seen on 3T pmMRI D. Details on multiple fetal malformations and skeletal dysplasias (Supplemental Table 2

Atrial septal defect
Postmortem T2-weighted 4-chamber view in a fetuses with an atrial septal defect type I in gestational week 23þ3. The defect represented by asterisk between the RA and LA can be visualized very well. In addition, the blood-fluid level is continuous through both atria.

Tetralogy of Fallot
Postmortem T2-weighted parasagittal slice of a fetus with tetralogy of Fallot in gestational week 36þ0. The overriding aorta represented by asterisk can be seen well.

SUPPLEMENTAL FIGURE 4
Hypoplastic right heart Postmortem T2-weighted 4-chamber view in a fetus with hypoplastic right heart syndrome in gestational week 23þ5. The arrow marks the hypoplastic right ventricle.

SUPPLEMENTAL FIGURE 6
Truncus arteriosus T2-weighted oblique-sagittal reconstruction of a fetus with truncus arteriosus communis in gestational week 26þ1. Both the ascending aorta, and the pulmonary trunk arise from a common trunk (arrow).

SUPPLEMENTAL FIGURE 8
Coarctation of aorta T2-weighted oblique-sagittal view (parallel to the aortic arch) of a fetus with coarctation of the aorta in gestational week 33þ0. The tapering in the proximal descending aorta and the coarctation (arrow) can be delineated well. The 3 asterisks mark the supraaortal branches (brachiocephalic trunk, left common carotid artery, and left subclavian artery).
Ao, descending aorta. Overview of autopsy findings in fetuses where multiple malformations or skeletal dysplasia are noted on Supplemental Table 2 Case ID from Supplemental Table 2 Main autopsy findings (summative).
15þ2, 37 g, IUFD Flexion deformity at the second and fourth fingers (bilateral) Values are expressed as median (range) or number (percentage).
3T pmMRI, 3Tesla postmortem magnetic resonance imaging; IUFD, intrauterine fetal death; NND, neonatal death. a In 21 cases, the maternal body mass index was unknown (21/220, 9.5%); b 11 fetuses or neonates were included in the 3T pmMRI study, and 7 healthy neonates were not included; c Included placental abruption and umbilical cord knot, without problems of twinning; d None of these cases exhibited a fetal or neonatal congenital heart defect at autopsy; e Included 1 dichorionic twin pregnancy with late miscarriage of both fetuses, 2 monochorionic twin pregnancies with intrauterine death of 1 twin and survival of the other, 2 monochorionic twin pregnancies with early twin-twin transfusion syndrome and demise of all 4 fetuses (3 included in the 3T pmMRI study), and 1 monochorionic twin pregnancy after laser treatment of twin-twin transfusion syndrome and demise of 1 twin.    Original Research OBSTETRICS ajog.org