A systematic autopsy survey of human infant bridging veins

In the first years of life, subdural haemorrhage (SDH) within the cranial cavity can occur through accidental and non-accidental mechanisms as well as from birth-related injury. This type of bleeding is the most common finding in victims of abusive head trauma (AHT). Historically, the most frequent cause of SDHs in infancy is suggested to be traumatic damage to bridging veins traversing from the brain to the dural membrane. However, several alternative hypotheses have been suggested for the cause and origin of subdural bleeding. It has also been suggested by some that bridging veins are too large to rupture through the forces associated with AHT. To date, there have been no systematic anatomical studies on infant bridging veins. During 43 neonatal, infant and young child post-mortem examinations, we have mapped the locations and numbers of bridging veins onto a 3D model of the surface of a representative infant brain. We have also recorded the in situ diameter of 79 bridging veins from two neonatal, one infant and two young children at post-mortem examination. Large numbers of veins, both distant from and directly entering the dural venous sinuses, were discovered travelling between the brain and dural membrane, with the mean number of veins per brain being 54.1 and the largest number recorded as 94. The mean diameter of the bridging veins was 0.93 mm, with measurements ranging from 0.05 to 3.07 mm. These data demonstrate that some veins are extremely small and subjectively, and they appear to be delicate. Characterisation of infant bridging veins will contribute to the current understanding of potential vascular sources of subdural bleeding and could also be used to further develop computational models of infant head injury.


Introduction
In early childhood, subdural haemorrhage (SDH) can occur through accidental and non-accidental mechanisms [1][2][3][4] as well as from birth-related trauma [5]. One of the proposed non-accidental mechanisms of injury involves a baby being subjected to repetitive rotational and acceleration/deceleration forces within the head due to violent shaking, with or without an additional impact force. The term abusive head trauma (AHT) is often used to include a pattern of injuries that are thought to occur through the inertial forces associated with a shaking event. Of these injuries, SDH is the most commonly observed feature and has been demonstrated in up to 92% of AHT cases [6].
The origin of infantile SDH has been debated throughout the scientific and medical literature [7][8][9]. Historically, damage to the bridging veins has been the most commonly suggested source of subdural bleeding in AHT [10][11][12]. However, previous experimental research on head injuries has centred around animal and biomechanical models [13,14]. These studies have been used to estimate the amount of force required to tear bridging veins (which some authors state are large blood vessels [15,16]) to magnitudes that would be far greater than that induced by inertial forces, such as those that can be produced by vigorous shaking. Direct evidence of ruptured bridging veins is also limited, demonstrated in only a few imaging studies by the extravasation of contrast agent [17] and during autopsy examination [18].
There are several detailed descriptions from neuroimaging and cadaver dissection studies on the locations, size and anatomical characteristics of adult bridging veins [19][20][21]. However, there is scanty published research on infant bridging veins, which would be expected to be proportionally smaller than adult vessels. The most comprehensive studies on the anatomical locations of adult bridging veins can be found in neurosurgical books and papers as it is crucial for neurosurgeons to be aware of these veins during various surgical approaches, as damage to a bridging vein can result in postoperative brain damage [22][23][24].
In this study, we map the number and anatomical locations of bridging veins from 43 human neonatal, infant and young child post-mortem examinations onto a 3D model of the surface of a representative infant brain. We also describe our observations of the drainage routes of these vessels, either directly into the dural venous sinuses or indirectly through venous pathways within the dural membrane. The mean diameter of 79 bridging veins from five additional cases within these age groups is also reported.

Case selection
Forty eight neonatal, infant and early childhood (up to 2 years of age) autopsies undertaken at Leicester Royal Infirmary between March 2014 and January 2017 as part of a regional paediatric autopsy service were included in this study (Table 1). Corrected gestational ages were recorded for neonates within the first week of life. The first 43 cases of the series were used for the 3D mapping of vein locations and total counts of vein numbers. Five subsequent cases were used for in situ measurement of bridging vein diameters (a 1-dayold male and a 1-day-old female, a 29-day-old female, an 18month-old male and a 2-year-old male). For all cases where in situ measurements were made of bridging veins, appropriate parental consent was given. Retrospective ethical approval was also obtained for the use of archived autopsy photographs of paediatric brains and bridging veins for research purposes (NRES 14/EM/0169).
Included in the series were three AHT cases and five perinatal head trauma cases (Table 1).

Abusive head trauma cases
Abusive head trauma cases were classified as such using the following criteria: (1) extra-axial haemorrhage (subdural and/or subarachnoid), (2) injury to the brain (hypoxic-ischaemic injury, parenchymal injury, diffuse a x o n a l i n j u r y, c e r e b r a l o e d e m a ) , ( 3 ) r e t i n a l haemorrhages, (4) other injuries consistent with shaking/shaking with an impact (metaphyseal fractures, pattern bruises, spinal injury), (5) inadequate history to explain the observed injuries or confessed shaking and (6) defendant convicted of murder/manslaughter during legal proceedings.
Autopsy images of the subdural haemorrhages present in the cases below are available in our previous publication [25]. Case numbers given below relate to the current Table 1.

Case 25
Case 25 was a 9-week-old male with thin film SDH ( Table 2). There were bruises to the head, face, chest, abdomen, back and right knee; old and fresh metaphyseal and corner fractures of the tibia, fibula and several ribs; subarachnoid haemorrhage over the surface of the brain; diffuse widespread ischaemia; focal and small amounts of βAPP deposition in the white matter of the cerebral hemisphere and brain stem structures; ischaemic changes in the spinal cord; severe retinal haemorrhages in the right eye and bilateral optic nerve sheath bleeding; and a single left eye retinal haemorrhage. The defendant was convicted of manslaughter.

Case 32
Case 32 was a 17-week-old male with thin film SDH ( Table 2). There was scalp bruising on the occiput with underlying skull fracture and damage to the skull sutures; subdural haemorrhage along the length of the spinal cord; subarachnoid haemorrhage over the brain and within the spinal cord; generalised hypoxic-ischaemic damage to the brain and axonal damage demonstrated by βAPP deposition; focal axonal damage to the spinal nerve roots; and bilateral and severe retinal haemorrhages. The defendant admitted to two separate occasions of shaking and was convicted of manslaughter.

Case 38
Case 38 was a 31-week-old female with thin film SDH ( Table 2). There was bruising over the vertex of the scalp; subarachnoid haemorrhage over the brain and spinal cord; SDH and extradural haemorrhage in the spinal cord; generalised hypoxic-ischaemic damage to the brain and axonal damage in ill-defined areas of the brain and in the spinal nerve roots; and extensive bilateral retinal haemorrhages. The defendant was convicted of manslaughter.   The term 'thin smear' was used to describe a transparent area of blood, a 'thin film' was an opaque collection of SDH which was not space occupying (≈ or ˂ 2-mm thick) and a 'thick' SDH was opaque and space occupying (˃ 2-mm thick) n/a infant older than 4 weeks of age Post-mortem visualisation of the bridging veins and subdural haemorrhage The calvarial bones were removed and the dura mater was optical cleared as previously described [25,26]. Before reflection of the dura, any SDH seen through the membrane, overlying the cerebral convexities, was photographically documented. The dura was carefully incised according to the diagram (Fig. 1a). The dural flaps were then slowly reflected to enable photography of the bridging veins over the surface of the hemispheres (Fig. 1b). The brain was then gently tilted and manipulated to reveal bridging veins within the interhemispheric fissure, at the frontal and temporal poles, down the sides of the temporal lobe and on the undersurface of the brain, including veins associated with the tentorium. After a thorough examination of all the areas of the right hemisphere for bridging veins entering the dura, the right hemisphere was removed and a similar examination of the left side of the brain was conducted. The tentorium was then incised (Fig. 1c) and slowly reflected to reveal the bridging veins draining from the cerebellum. After the tentorium was lifted, the cerebellum was carefully manipulated to view any bridging veins remaining on the lateral and inferior surfaces. Vessels associated with the cranial nerves were not included. Subdural haemorrhage was further documented on reflection of the dura, after removal of each hemisphere, on reflection of the tentorium and after removal of the cerebellum. This allowed the presence of interhemispheric, supratentorial and subtentorial SDH to be recorded, as well as blood within the cranial fossae.

Magnetic resonance imaging of an infant brain
For the purposes of creating a 3D model of an infant brain, magnetic resonance imaging (MRI) scans were undertaken on a brain from an infant with an age (11 weeks) close to the median of the case series, with specific parental consent. The brain was removed as part of the standard autopsy procedure and was placed in 20% formalin for approximately 2 weeks before scanning to allow for adequate fixation of the tissues. The brain was suspended in a specifically designed plastic container with the aid of two plastic rods to ensure that the organ was surrounded by fluid. This facilitated subsequent separation of the brain from the walls of the container on the MRI scans. Both T1-weighted and T2-weighted spin echo sequences were used to image the brain.
Segmentation of the brain using ITK-SNAP Using the T2-weighted MRI images (as the edges of the brain appeared to be more distinct in this sequence), the two hemispheres and the cerebellum were digitally isolated from each other by a combination of semi-automatic and manual segmentation using an open-source multi-platform software application (ITK-SNAP, version 3.2.0) [27] which is designed to segment structures in 3D medical images. Each part of the brain (hemispheres and cerebellum) was initially semiautomatically segmented, by outlining a 3D point of interest and by applying an algorithm to this area based on pixel intensity and image edges. Further manual segmentation was then required due to similar pixel intensities in the areas of interest and because the algorithm was unable to detect two small sections on the lateral aspects of the hemispheres due to reduced quality of the MRI images in those areas. Each segmented area of the brain was given a numerical label. The 3D brain was exported from ITK-SNAP as a mesh stl file. The numerical labels enabled the three separate components of the brain to be recognised by Matlab, a numerical computing environment.
Creation of a Matlab Interface for plotting bridging vein locations on a 3D infant brain model A user interface was developed using a commercial software package (MATLAB 2015a, The MathWorks, Inc., Natick, Fig. 1 Incising and reflecting the dural membrane. a Incision lines along the right coronal and lambdoid sutures. b Reflection of the right frontal dura mater. c Incision lines on the tentorium MA, USA) to enable the post-mortem photographic data to be mapped onto the 3D infant brain model produced in ITK-SNAP. The interface was created so that the brain could be viewed as its three separate components (hemispheres and cerebellum). This enabled plotting and visualisation of data points that would otherwise be hidden if viewing the brain as an entire organ (i.e. the interhemispheric fissure and the tentorium). To allow for data point plotting, the interface could rotate both the whole 3D brain and the separate hemispheres/ cerebellum and also had the ability to save and load separate cases and groups of cases. A heat map function was added using 'knnsearch' and 'rangesearch' algorithms. The number of bridging veins on each of the hemispheres and the cerebellum were also recorded by the interface, allowing for the average number of vessels to be calculated for each of the separate parts and also for the whole brain.

Macrophotography and measurement of bridging veins
In five post-mortems, in situ measurements of the diameter of 79 bridging veins were recorded. The dura was reflected and easily accessible bridging veins (on the cerebral convexities including the parasagittal sections of the frontal and parietal lobes, within the interhemispheric fissure and on the superior surface of the cerebellum) were captured using a digital SLR camera. A blue plastic-coated wire of known diameter (0.22 mm) was placed adjacent to the each bridging vein and used as a coplanar scale reference (Fig. 2). Vessel diameters at the mid-point of the bridging section (the part of the vein which left the pia/arachnoid and joined either the dura mater or sinuses directly) were calculated using ImageJ image analysis software (National Institutes of Health, USA) from the pixel dimensions of the blue wire. Only singular vessels which were not obviously seen to branch within the bridging section of the vein were photographed for measuring purposes, due to the more easily defined edges of non-branched isolated blood vessels compared to branching vessels in close approximation.

Post-mortem visualisation of the bridging veins and subdural haemorrhage
We systematically located and photographically documented the macroscopically visible bridging veins in all 43 cases. The vessels were typically observed in the parasagittal region close to the superior sagittal sinus (SSS), including vessels which appeared to be directly entering the sinus (Fig. 3a). Further common bridging vein locations included at the frontal and temporal poles (Fig. 3b, c), the inferior aspect of the posterior temporal lobe/occipital lobe (Fig. 3d) and on the superior surface of the cerebellum (Fig. 3e). Although bridging veins were most often seen close to the venous sinuses, a few were also seen on the cerebral convexities distal to the sinuses and these vessels were often extremely small and could be missed when viewed with the naked eye (Fig. 3f). Veins that did not directly enter the more well-known, relatively large, dural venous sinuses, appeared to enter smaller venous channels on the inner surface of the dura (Fig. 3g), often in the parasagittal region, extending 2 to 3 cm from the midline (Fig. 4a, b) or within the tentorium (Fig. 4c). These venous channels had a flattened appearance, either within the inner aspect of the membrane itself or occasionally appearing to have a loose attachment to the membrane. In the three AHT cases, the bridging veins tended to appear more obviously engorged with congested blood, particularly the larger vessels in the parasagittal region, and on application of slight pressure to the vessels, they did not blanch (Fig. 3h) (in a small study of a subset of 15 sequential non-head injured cases only 6 had blood vessels which did not blanch and of these cases only 1-3 veins in each case were observed to blanch under slight pressure).
In addition to the eight AHT and perinatal head injury cases, a further five neonates and infants also had SDH ( Table 2). Of these five cases, only one infant was over 4 weeks of age, and this was a 14-week-old male with minor crush injury to the head due to overlaying. The three AHT  No other cases demonstrated this particular pattern, but the appearance of SDHs in other cases ranged from small focal thin smears of blood to extensive space occupying haemorrhage.

3D mapping of the bridging veins
Using the 'knnsearch' and 'rangesearch' algorithms to produce location density maps of the bridging veins, we identified large numbers of vessels near the frontal and parietal sections of the SSS, at the frontal and temporal poles, on the inferior aspect of the temporal and occipital lobes following close to the transverse sinus and around the superior vermis and the anterosuperior border of the cerebellar hemispheres (Fig. 5).
Although not demonstrated on the heat maps due to their smaller numbers and therefore smaller densities, when viewing the raw data points on the brain, bridging veins were also occasionally noted in locations situated further away from the sinuses during the post-mortem examinations (approximately 4%) (Fig. 6). The five lowest counts for total bridging veins were seen in the three AHT cases included in the series (27, 30 and 34 veins) and two cases of perinatal head trauma (30 and 33 veins), born respectively by forceps and ventouse delivery methods (Tables 1 and 2). All five of these cases had SDH covering a significant amount of the cerebral surface and within the posterior fossa.

Outer diameter of infant bridging veins
The mean diameter of the 79 bridging veins was 0.93 mm (range 0.05-3.07 mm). Due to the clustering and accessibility of bridging veins close to the SSS, most of the diameter measurements were taken from parasagittal vessels. In total, 37 bridging veins were recorded on the right hemisphere as opposed to 28 on the left hemisphere. Bridging veins in the interhemispheric fissure and cerebellum were also measured ( Table 3).

Discussion
Post-mortem studies of the infant bridging veins are technically challenging due to their location below the dural membrane, as the infant dura is still attached to the calvarium at the sutures, rendering removal of the calvarium without damage to the bridging veins extremely difficult by conventional methods. When inspected at autopsy, bridging veins are fragile and easily disrupted by applying excessive stretching force during manipulation and under the force of gravity. Paediatric  Light microscopy, X-ray, digital subtraction angiography and cadaveric anatomical dissections with vascular casting have previously been used to determine the diameter, number and anatomical locations of bridging veins [28][29][30]. Variation exists in the diameter measurements reported in these studies, with values ranging between 1 and 7 mm, except when using anatomical dissections where veins as small as 0.1 mm have been infrequently recorded [19]. In a radiological study, the diameters were shown to vary considerably between individuals but not in any one individual [28]. To our knowledge, no studies have specifically detailed the physical characteristics of infant bridging veins, with the exception of one study in which outer diameter measurements of between 0.14 and 0.72 mm were recorded for six vessels [31].
Owing to the delicate nature of the bridging veins, we were only able to measure the diameters of the larger bridging veins in the more accessible areas (i.e. the frontoparietal parasagittal regions and superior cerebellum). Due to the collapse of the blood vessels and lack of blood flow, post-mortem measurements can only be considered to represent approximations of the true in vivo vessel diameters. However, it is likely that the stated values for diameter measurements are over-estimations due to the flattened appearance of the veins from collapse and also due to the necessity of photographically documenting the veins lying against a very thin wire or probe to ascertain the bridging section for subsequent measurement of vein diameters. Although more bridging veins were measured from the right hemisphere than the left, it is possible that this is due to our standard method for opening the dura overlying the right hemisphere first, rather than reflecting anatomical asymmetry in these cases. Despite these limitations, the reported values were still considerably smaller than those documented in the majority of adult studies [29,30] but broadly correlate with the limited measurements recorded by Morison for infant bridging veins [31].
There is considerable variation in the number of bridging veins recorded between studies in the literature [19,20,30], with average values ranging from 29 veins in one study [19] to only four or five in another study [20]. However, these studies often refer only to the larger veins draining directly into the SSS. Differences in reported numbers may also result from the various methods of investigation, cohort numbers and ages of the participants. A further area of study for bridging vein locations is underneath the tentorium, where vessels often enter the left and right transverse sinuses [32][33][34][35]. Neurosurgical papers also provide descriptions of bridging veins when determining the safety of several surgical approaches. However, these publications are usually focused on small, specific areas, often focused on the drainage of blood into a specific dural venous sinus [32,36]. To our knowledge, this paper is the first systematic paediatric autopsy study to map all the bridging veins visible with the naked eye, in any one individual, over the entire surface of the brain. For this reason, our reported average number of infant bridging veins is considerably higher than values reported previously in studies in adult populations. It is probable that our average bridging vein count may nevertheless represent an underestimation owing to the technical difficulties of examining very small veins, documenting vessels located in cranial recesses with limited visibility and the necessity to incise the dura to gain access. However, we consider that our counts represent a best estimate of bridging vein numbers that could be identifiable by the naked eye under otherwise ordinary autopsy conditions. We have also demonstrated the considerable variation in number of bridging veins between individuals. Although our counts of bridging veins were lower in the AHT cases than the non-head injured cases, we cannot at this point determine whether this was due to the elastic recoil of small calibre broken veins, decreasing the likelihood of detection with the naked eye, or by high numbers of veins being obscured from view by the presence of SDH.
Our technique of using neurosurgical equipment and meticulous dissection methods have allowed us to demonstrate the anatomical locations of infant bridging veins, including vessels which 'bridge' to the dura distant from the sinuses and therefore demonstrating indirect routes of drainage into the larger dural venous sinuses. The locations of bridging veins recorded within this paper are comparable with previous publications [19,21,37], and although variations in the nomenclature used to describe these vessels exist in the published literature, they have previously been grouped as: a superior sagittal group, a sphenoidal group, a tentorial group and a falcine group, depending on which dural venous sinus they drain into [38].
Tentorial sinuses (sometimes described as meningeal veins or lakes) have been well described in the literature as constant venous channels [35,37,39]. Bridging veins from both the cerebellum and cerebrum may enter the tentorial sinuses which then drain into the transverse and straight sinuses including at the point of the transverse-sigmoid junction [40]. However, we observe that bridging veins also appeared to join venous channels either loosely attached to the dura or within the membrane itself, often in the parasagittal regions.
Although not often alluded to in the literature, this indirect drainage route to the sinuses has been noted previously by some authors who describe 'adhesions' of the bridging veins to the dura [19] and venous sinuses (sometimes referred to as meningeal veins) [30,38,41] extending out by more than a centimetre from the SSS. In several more historical papers, as well as being described as either attached or within the dura, bridging veins are suggested to be free in the subdural space for up to 3 cm [42,43], which one author describes as being particularly noticeable in the infant [42]. Through microscopic examination, the subdural space has been shown to be a potential space as the dura and arachnoid are continuous with one another [44]. A space is suggested to be created by the collection of fluid or air which cleaves through a weak cell layer of the dura [45] known as the dural border cell layer [46]. During our observations of the bridging veins, upon reflection of the dura, we create a subdural space; therefore, we cannot say how much of the free bridging sections of the veins were or were not within the dura. It is possible that the veins which appeared loosely attached to the dura may have occupied the dural border cell layer, disrupted by reflection of the membrane.
As bridging veins are the most commonly suggested source of infantile SDH, detailed descriptions of these blood vessels in the young are highly important for medical professionals involved in infantile head injury cases. We have shown that not all bridging veins are large calibre vessels and that some vessels are extremely small. They also appear to be delicate as they are easily disrupted on reflection of the dura.
As shown within this paper and others [20,32,37], the cerebral drainage system is extremely variable from one individual to the next. Whether this variation in vasculature could result in one infant being more susceptible to traumatic head injury and subdural bleeding than another cannot be answered here, but this point is worth future consideration, especially given the wide range of clinical outcomes seen in AHT cases. Anatomical information, including the data within this paper, may aid future understanding of the potential mechanisms which result in SDH and confirm whether or not bridging veins can be traumatically damaged by inertial/rotational forces to produce this type of bleeding. These data could also provide biomechanical engineers with the opportunity to create more detailed computational models of human infantile head trauma, accidental or otherwise.