Birth-related subdural hemorrhage in asymptomatic neonates: evolution over time and differentiation from traumatic subdural hemorrhage

Abstract

Background:

Intracranial birth-related subdural hemorrhage frequently occurs in asymptomatic newborns and has no adverse long term sequelae. It is medically and medicolegally important to differentiate birth-related subdural hemorrhage from other pathological causes of intracranial hemorrhage. There is limited literature available on the incidence of birth-related subdural hemorrhage, its imaging features and evolution over time, mainly because asymptomatic infants do not routinely undergo cranial MRI.

Objectives:

To establish the incidence and distribution of birth-related subdural hemorrhage and evaluate their association with various modes of delivery, identify associated features and evaluate the sequential evolution of signal changes of the birth related hemorrhages on MRI.

Material and Methods:

A total of 200 healthy term neonates and young infants were included in this retrospective review study. All infants underwent MRI of the brain and cervical spine at postnatal age of 0–2 months with acquisition of a 3D T1-weighted (T1W), 3D or 2D T2-weighted (T2W) and axial diffusion-weighted imaging (DWI) sequences. The scans were evaluated for the presence and distribution of subdural hemorrhages, other intracranial hemorrhages and associated injuries. Prevalence of intracranial hemorrhage in various modes of delivery was analyzed. Relationship between the signal intensities of the bleeds on T1W, T2W and DWI scans and the age of the infants was analyzed. Appropriate tests were applied to test for statistical significance of the data.

Results:

Out of 200 neonates, 66 (33%) had detectable intracranial hemorrhage on MRI with an age range of 11–25 days, including 31 (47%) males and 35 (53%) females. All of them had subdural hemorrhages, 54 (81.8%) of which were in the posterior fossa. Additional parenchymal hemorrhages were present in a few, but no subarachnoid hemorrhages, cervical spinal canal hemorrhages, cortical bridging vein injury or cervical spinal ligamentous injury were identified within the limitations of the study. No detectable intracranial hemorrhage was found in subjects above 25 days of age. Overall incidence of subdural hemoeehage by mode of delivery was 8/68 (11.8%) in babies born by cesarean section and 58/132 (43.9%) in babies born by vaginal delivery. Among the vaginal deliveries, the highest incidence was observed in assisted vaginal delivery (19/30, 63.3%). Subjects with birth-related subdural hemorrhage were categorized into three age groups: <13 days, 13–21 days and >21 days. All detected hemorrhages were T1W hyperintense. In the <13 days group, all bleeds were T2W hypointense. In the 13–21 days group, 73.1% were T2W hypointense, while 26.9% were T2W mixed. All bleeds in the >21 days group were T2W hypointense. All DWI hyperintense bleeds were found in the 13–21 days group.

Conclusion:

Birth-related subdural hemorrhage occurs in over a third of normal deliveries and has a characteristic distribution, predominantly in the posterior fossa. Associated cervical spinal subdural hemorrhages, cervical spinal ligamentous injury or cortical bridging vein injury, which are concerning for traumatic etiology, were not identified. Birth-related subdural hemorrhages follow a characteristic pattern of signal changes on MRI. Although not completely reliable, this can help in differentiating them from traumatic intracranial hemorrhages which usually occur postnatally. No birth related subdural hemorrhages were seen after 25 days of age in our cohort.

Introduction:

Birth trauma is estimated to account for 1–2 % mortality in the neonatal period [1–3]. Intracranial hemorrhages are a common component of birth related injuries. Symptomatic intracranial hemorrhages are an important cause of neonatal mortality [4]. However, intracranial hemorrhage also frequently occurs in asymptomatic newborns. These are usually subdural hemorrhages that are small in size and are commonly referred to as birth-related subdural hemorrhages, as such hemorrhages are extremely common and without any clinical symptoms or sequelae [5]. Data on incidence of birth-related subdural hemorrhage is scant, but has been estimated to be 26–63 % on the basis of various studies [2, 5]. While these small bleeds may be visualized on computed tomography (CT), in view of radiation concerns, magnetic resonance imaging (MRI) is the preferred modality. There is limited literature available on the incidence of birth-related subdural hemorrhage, its imaging features and evolution over time, mainly because asymptomatic infants do not routinely undergo cranial MRI. These bleeds are often difficult to visualize on cranial ultrasound which may be performed in some infants for screening. The importance of analyzing such birth-related subdural hemorrhage in asymptomatic infants is in differentiating this benign entity from traumatic intracranial hemorrhages, especially those related to non-accidental trauma in early infancy. This is often challenging in the neonatal period as traumatic intracranial hemorrhage is also most likely to be in the subdural location [5, 6]. The medical implications from a clinical management standpoint, as well as medicolegal implications in cases of non-accidental trauma, necessitate differentiation of birth-related subdural hemorrhage from traumatic causes of intracranial hemorrhage.

With this background, we performed this study by imaging asymptomatic neonates and young infants using MRI with several objectives. We aimed to establish the incidence of subdural hemorrhage in asymptomatic term infants, understand the distribution of birth-related subdural hemorrhage in asymptomatic infants, examine the association of birth-related subdural hemorrhage with various modes of delivery, with cervical spinal subdural hemorrhage and with pathological features such as cervical spinal ligamentous injury, parenchymal injury and cortical vein injury (lollipop sign). Additionally, we aimed to evaluate the duration of persistence of benign birth-related subdural hemorrhage in the posterior fossa. We also attempted to time the bleed based on appearance of the blood on MRI, assuming that all the bleeds occurred at birth.

Material and Methods:

A total of 200 healthy term neonates and young infants (born between 37 and 42 completed weeks of gestation) were included in this retrospective review study. All subjects had birth weight ranging from 2.5 kg to 4 kg and did not require resuscitation, hospital admission or special care after birth up to the time of the scan. These subjects were participants of two ongoing research projects on asymptomatic birth-related intracranial hemorrhage, both of which involved an MRI examination of the brain performed during the first 2 months of life on randomly selected healthy asymptomatic subjects born to healthy pregnant women without any complications during pregnancy (clinical trial number NCT04274140). Exclusion criteria were identifiable causes or risk factors for intracranial hemorrhage including head trauma, non-accidental injury and prematurity, as well as additional risk factors such as known coagulopathies. Institutional review board (IRB) approval was obtained for all study activities prior to starting the studies, and written and informed consent was obtained from the parents. The mode of delivery information was obtained from the two research studies and subjects were included into one of four groups, namely vaginal delivery, assisted vaginal delivery, routine cesarean section and emergency cesarean section. All infants underwent MRI of the brain between 0–2 months of age using either a Siemens PRISMA 3T MRI (Siemens Healthineers, Erlangan, Germany) or a Philips Achieva 1.5T MRI (Philips, Amsterdam, Netherlands) machine. A 3D T1W sequence with multiplanar reconstructions, either a 3D T2W (Siemens 3T) fat-suppressed sequence with multiplanar reconstructions or a 2D T2W (Philips 1.5T) fat-suppressed sequence in all three planes and an axial diffusion weighted imaging (DWI) sequence with b-value 1000 were acquired. The detailed scanning protocol used is summarized in Supplementary Material 1. On the T1W and T2W scans, the field of view was extended to include the entire cervical spine up to the lower border of C7 vertebra, so that T1W and T2W scans of the cervical spine were available in all three planes for evaluation in all subjects. All scans were acquired during natural sleep without sedation. As part of the research projects, the MRI scans were read by neuroradiologists to screen for incidental findings.

The scans were independently read to evaluate for presence of or absence of subdural hemorrhages by a fellowship trained pediatric neuroradiologist with over 15 years’ experience in the field and two pediatric radiology fellows with over 5 years’ experience in general radiology. Distribution of the subdural hemorrhages was categorized as within the posterior fossa or supratentorial. Presence of hemorrhage along the tentorium was also noted. Within the supratentorial compartment, the location of the hemorrhage was subcategorized as frontal, temporal or parieto-occipital or a combination of these. Presence of any associated or isolated cervical spinal canal subdural hemorrhage was also noted. The scans were also evaluated for other intracranial hemorrhages such as parenchymal hemorrhages and subarachnoid hemorrhages. The maximum width of the subdural hemorrhages was measured and the largest width of the hemorrhage in each subject was documented. Maximum dimension of the parenchymal hemorrhages was also noted. Signal intensity of all hemorrhages on T1W and T2W sequences were documented. In cases with multiple subdural hemorrhages of varying signal intensities, the signal intensity pattern of the dominant hemorrhage was documented. Additional pathological features such as cervical spinal ligamentous injury, parenchymal injury and cortical vein injury (lollipop sign) or thrombosis, if any, were also documented. All cases had diagnostic quality T1W scans. Cases with non-diagnostic T2W or DWI sequences were included in the study, but excluded from any statistical analysis specifically requiring the signal intensity of the bleed on these sequences. A total of 2 cases had non-diagnostic T2W sequences and 2 cases had non-diagnostic DWI sequences.

Statistical analysis of the collected data was performed. The age and gender distribution of the subjects and the prevalence of subdural hemorrhage as well as their relationship with the age of the subjects was calculated. Prevalence of intracranial hemorrhage in various modes of delivery was analyzed using the chi squared test. Distribution of the subdural hemorrhages within the intracranial compartments was also evaluated. The neonates with birth-related subdural hemorrhage were divided into three categories based on age as <13 days, 13–21 days and >21 days of age. The selection of the age groups was based on the signal characteristics of the bleeds, since all subjects with mixed T2W signal and with hyperintense DWI signal of the bleeds were between 13 and 21 days of age. This age group also included a majority of the subjects. The remaining subjects were hence categorized into either younger than 13 days or older than 21 days of age. The blood signal intensities on T1W, T2W and DWI scans were assessed and the relationship between the signal intensities of the bleeds (assumed to be birth-related) and the age of the neonates was analyzed. Agreement statistics between the reviewers was analyzed and concordance was calculated using Cohen’s kappa. Statistical analysis of the data was performed using appropriate statistical tests on Matlab software (Version R2018b, MathWorks Inc., Natick, Massachusetts, USA), with P < 0.05 considered significant.

Results:

A total of 200 asymptomatic term neonates and young infants were imaged in the study with an age range of 9–54 days. The mean age and standard deviation for all participants was 18 ± 5.7 days, with a median age of 17 days (Figure 1). A total of 106/200 (53%) infants were males and 94/200 (47%) were females. Of the 200 infants, 68 were born by cesarean section and 132 were born by vaginal delivery. The demographic data of the study population is summarized in Table 1. Intracranial hemorrhage was detectable on MRI in 66/200 (33%), of which 31 (47%) of these were males and 35 (53%) were females, without any statistically significant difference in incidence of ICH between males and females. Subjects with hemorrhage on MRI were neonates in the age range of 11–25 days. The mean age and standard deviation for neonates with hemorrhage was 16.5 ± 3.7 days, with a median age of 16 days (Figure 1). All (100%) neonates with intracranial hemorrhage had subdural hemorrhages. In addition to subdural hemorrhage, 8/66 (12.1%) had additional parenchymal hemorrhages. None of them had subarachnoid hemorrhages. There were also no infants with visible subdural hemorrhages in the visualized cervical spine.

Fig. 1.

a Histogram representing distribution of number of subjects by postnatal age at the time of the magnetic resonance imaging (MRI) scan. b Histogram representing distribution of number of subjects with subdural hemorrhage by postnatal age overlaid on distribution of number of subjects by postnatal age

Table 1.

Demographic data of the study subjects

MRI scanner (Sample size)	1.5 T (n = 110)	3.0 T (n = 90)	Total (n = 200)
Gender (Male/Female)	61 / 49	45 / 45	106 / 94
Postnatal age at MRI (Days) Mean ± SD (Range)	17.1 ± 4.0 (9 to 31)	19.1 ± 7.0 (10 to 54)	16.5 ± 3.7 (9 to 54)
Delivery type (Routine Vaginal/Assisted Vaginal/Routine Caesarean section/Emergency Caesarean section)	45 / 20 / 28 / 17	57 / 10 / 17 / 6	102 / 30 / 45 / 23

The majority of the subdural hemorrhages (54/66, 81.8%) were located in the posterior fossa (Figure 2). Subdural hemorrhages in the supratentorial compartment were present in 28/66 (42.4%) (Figure 3). Subdural bleeds along the tentorium were found in 15/66 (22.7%), which were also included in the posterior fossa or supratentorial compartment or both, based on the surface of the tentorium involved (Figure 3). Among the 28 cases with subdural hemorrhages in the supratentorial compartment, 7/28 (25%) had bleeds only on the left side, 2/28 (7.1%) had bleeds only on the right side and 19/28 (67.9%) had bleeds bilaterally. Bleeds in the parieto-occipital region including the posterior interhemispheric region were present in 27/28 (96.4%) (Figure 3). In addition to parieto-occipital subdural hemorrhage, additional bleeds in the frontal region were present in 2/28 (7.1%) and additional bleeds in the temporal region were present in 2/28 (7.1%). Only one case (3.6%) had isolated subdural hemorrhage in the temporal region without associated parieto-occipital hemorrhage, meaning a total of 3/28 (10.7%) had bleeds in the temporal region. Among the 8 cases with parenchymal hemorrhages in addition to subdural hemorrhage, all 8 had hemorrhages in the parieto-occipital cerebral parenchyma (Figure 4). There were no parenchymal hemorrhages in the posterior fossa. The types and distribution of hemorrhages in the study subjects is summarized in the flowchart in Figure 5. None of the cases had associated injury to the cervical spinal ligaments. Cortical vein thrombosis or evidence of venous injury (lollipop sign) was not present in any of the neonates.

Fig 2.

17-day-old asymptomatic male neonate born through routine vaginal delivery. a Non-contrast axial T1-weighted image of the brain shows posterior fossa subdural hemorrhages which appear hyperintense (arrows). b Axial T2-weighted image of the brain shows the subdural hemorrhages to be mixed in signal (arrows)

Fig 3.

14-day-old asymptomatic female neonate born through assisted vaginal delivery. a Non-contrast sagittal T1-weighted image of the brain shows subdural hemorrhages in the posterior fossa and supratentorial region extending along the tentorium which appear hyperintense (arrows). b Sagittal T2-weighted image of the brain shows the subdural hemorrhages to be hypointense (arrows)

Fig 4.

14-day-old asymptomatic male neonate born through routine vaginal delivery. a Non-contrast coronal T1-weighted image of the brain shows two tiny parenchymal hemorrhages in the left parieto-occipital parenchyma in the periventricular region which appear hyperintense (arrow). b Axial T2-weighted image of the brain shows the parenchymal hemorrhages to be hypointense (arrow)

Fig 5.

Flowchart depicting the types and distribution of intracranial hemorrhages in study subjects with a total study sample size of 200. Some of the subjects had multiple types of hemorrhages and/or hemorrhages in multiple locations, and these have been included in each subcategory.

The maximum width of the subdural hemorrhages ranged between 1 and 5 mm. A total of 63/66 (95.5%) of the subjects had subdural hemorrhages measuring 3 mm or less. Two subjects had subdural hemorrhages measuring 4 mm and one subject had subdural hemorrhage measuring 5 mm in width. The mean and standard deviation of the maximum width of subdural hemorrhages was 1.8 ± 0.9 mm with a median width of 2 mm. The maximum dimension of all 8 parenchymal hemorrhages ranged between 1 and 2 mm in maximum dimension. No parenchymal hemorrhage was identified measuring more than 2 mm in size.

Overall incidence of subdural hemorrhage by mode of delivery was 8/68 (11.8%) in babies born by cesarean section and 58/132 (43.9%) in babies born by vaginal delivery. Incidence of asymptomatic subdural hemorrhage according to the mode of delivery was highest in assisted vaginal delivery which was seen in 19/30 (63.3%) cases. Subdural hemorrhage in routine vaginal delivery was seen in 39/102 (38.2%) cases. A total of 4/45 (8.9%) cases of routine cesarean section and 4/23 (17.4%) cases of emergency cesarean section showed presence of asymptomatic subdural hemorrhages. The occurrence of intracranial hemorrhage by type of delivery in the study subjects is summarized in Table 2. The incidence of intracranial hemorrhage in vaginal delivery was significantly higher than the incidence of intracranial hemorrhage in cesarean section (P < 0.05). The incidence of intracranial hemorrhage in assisted vaginal delivery was also significantly higher than incidence of intracranial hemorrhage in normal vaginal delivery (P < 0.05). The difference in incidence of intracranial hemorrhage between emergency and routine cesarean section was not statistically significant.

Table 2.

Occurrence of intracranial hemorrhage by type of delivery in the study subjects

Type of delivery	Occurrence of hemorrhage
Routine vaginal delivery	39 of 102 (38.2%)
Assisted vaginal delivery	19 of 30 (63.3%)
Routine Caesarean section	4 of 45 (8.9%)
Emergency Caesarean section	4 of 23 (17.4%)

All 66 hemorrhages detectable on MRI were homogeneously hyperintense on T1W scan. There were differences in the T2W signal of the bleeds. The T2W sequence was not interpretable on 2 of the subjects with hemorrhage due to excessive motion, and these subjects were excluded from the analysis of T2W signal characteristics. Out of the remaining 64 subjects with bleeds, a total of 50 (78.1%) were T2W homogeneously hypointense (Figure 3), 14 (21.9%) were of mixed signal intensity (Figure 2) on T2W and none were T2W homogeneously hyperintense. The oldest neonate showing detectable hemorrhage was 25 days of age. A total of 7/66 (10.6%) neonates with subdural hemorrhage were <13 days, 54/66 (81.8%) were 13–21 days and 5/66 (7.6%) were >21 days of age. The T2W signal intensity of the bleeds when correlated with the age of the neonates revealed that all 7 hemorrhages in the <13 days’ group were hypointense on T2W scans. Hemorrhages in the 13–21 days’ group were variable in T2W signal and were hypointense or mixed. Out of the 52 subjects in the 13–21 days’ age group with interpretable T2W scans, a total of 38 (73.1%) were T2W hypointense while 14 (26.9%) were of mixed signal intensity on T2W. All 5 bleeds in the >21 days’ group were hypointense on T2W. All 14 mixed T2W signal hemorrhages were in the 13–21 days’ group (Figure 6). The DWI sequence was not interpretable on 2 of the subjects with hemorrhage due to excessive motion, and these subjects were excluded from the analysis of DWI signal characteristics. Out of the remaining 64 subjects with bleeds, 45/64 (70.3%) of the subdural bleeds were hypointense while 19/64 (29.7%) were hyperintense on DWI (Figure 7). All DWI hyperintense bleeds were in the 13–21 days’ group (Figure 6). None of the cases in the <13 days’ group or >21 days’ group showed hyperintensity on DWI. DWI hypointense bleeds were observed in all three age groups. The signal patterns of hemorrhage by age group in study subjects with intracranial hemorrhage is summarized in Table 3.

Fig 6.

a Bar chart showing distribution of T2-weighted imaging (T2W) signal of subdural bleeds in relation to age of the neonate. b Bar chart showing distribution of diffusion-weighted imaging (DWI) signal of subdural bleeds in relation to the age of the neonate

Fig 7.

15-day-old asymptomatic female neonate born through routine vaginal delivery. a Non-contrast sagittal T1-weighted image of the brain shows subdural hemorrhage in the posterior fossa which appears hyperintense on T1W (arrow). b Axial diffusion-weighted image of the brain with a b-value of 1000 shows the subdural hemorrhage to be hyperintense (arrow). c Axial apparent diffusion coefficient (ADC) map of the brain shows the subdural hemorrhage to be hypointense (arrow) consistent with restricted diffusion

Table 3.

Signal patterns of hemorrhage by age group in study subjects with intracranial hemorrhage

Age group	T1W	T2W	DWI
< 13 days (n = 7)	All hyperintense	All hypointense	All hypointense
13 – 21 days (n = 54)	All hyperintense	38 (73.1%) hypointense 14 (26.9%) mixed signal 2 non-interpretable	33 (63.5%) hypointense 19 (36.5%) hyperintense 2 non-interpretable
> 21 days (n = 5)	All hyperintense	All hypointense	All hypointense

Regarding the presence of hemorrhage, the pediatric neuroradiologist and the fellows were in agreement in 98.5% of cases. A subtle bleed was missed by the fellows which was identified by the pediatric neuroradiologist. Regarding the T2 signal of the bleeds, the readers were in agreement in 97% of cases. In the cases with inter-reader discrepancies, consensus opinion was considered and the opinion of the pediatric neuroradiologist after group discussion was considered final. Concordance testing between the reviewers using Cohen’s kappa for interpretation of presence and signal intensities of the bleeds showed a Cohen’s kappa value of 0.802 consistent with substantial agreement (P < 0.05).

Discussion:

Birth-related intracranial hemorrhage is a well-known entity. However, there is a relative paucity of literature on this entity and the available studies also have small sample sizes, mainly because asymptomatic neonates do not usually get imaged in the first month of life. Any clinical indication for imaging tends to confound the evaluation of asymptomatic intracranial hemorrhage secondary to the birth process. Additionally, most cranial imaging in the first few months of life is performed using ultrasound. MRI is limited to neonates with significant findings detected on ultrasound or clinically symptomatic neonates. Ultrasound is often unable to detect the subtle birth-related subdural hemorrhages. However, detailed understanding of the prevalence, distribution and sequential evolution of birth-related ICH, especially subdural hemorrhages, is very important due to the need for differentiation of these birth-related hemorrhages from traumatic intracranial hemorrhages. This is of particular significance in cases of intracranial hemorrhage with suspected abusive head trauma that are also usually subdural in location [6], and identification of subdural hemorrhage as being secondary to abusive head trauma rather than birth-related is both clinically and medicolegally important.

Birth-related subdural hemorrhage and mode of delivery:

A recently proposed theory for occurrence of birth-related subdural hemorrhage is the presence of relatively fragile prominent dural venous plexuses in the fetal dura between the periosteal and meningeal layers of the neonatal brain. Birth-related subdural hemorrhage tends to occur along the posterior falx, tentorium and in the posterior fossa, matching the sites of these prominent dural venous plexuses and lending credence to this theory [2, 7]. However, earlier studies had established a number of etiologies relating to birth trauma with falcine and tentorial tears, tears involving the dural venous sinuses, vermian veins or vein of Galen or occipital diastasis during birth [1, 5]. The actual hemorrhage has been proposed to occur due to the birthing process secondary to stretching and mechanical compression of these veins during passage through the birth canal [8]. Accordingly, some studies have found increased occurrence of subdural hemorrhage with vaginal delivery [2, 5, 9, 10]. Some other studies have found that complicated delivery is actually more important than the mode of delivery in determining the incidence of asymptomatic subdural hemorrhage. Towner et al. found that instrumental vaginal delivery and cesarean section carried a much higher risk of asymptomatic subdural hemorrhage than normal vaginal delivery [11]. Some association has also been reported between the presence of cephalhematoma and subdural hemorrhage [5]. The results of our study were similar to existing literature, with 43.9% of babies born by vaginal delivery showing presence of asymptomatic subdural hemorrhage as against 11.8% of babies born by cesarean section. Additionally, when considering only assisted vaginal deliveries, the incidence of asymptomatic subdural hemorrhage rose to 63.3% in our study.

Patterns and distribution:

The overall incidence of intracranial hemorrhage in our study was 33%. This is comparable with prior studies [5, 12, 13]. All neonates with intracranial hemorrhage had subdural hemorrhage with a few showing additional parenchymal hemorrhages. None of them had subarachnoid hemorrhage. Studies have found subarachnoid hemorrhage to be relatively less common with birth-related intracranial hemorrhage in asymptomatic neonates [5, 13]. Animal studies have shown that subarachnoid hemorrhage, even if small and isolated, is probably more likely to be symptomatic since blood is highly irritant to the leptomeningeal lining [14]. Recent molecular studies on animal brains have shown that subarachnoid hemorrhage disrupts normal brain development, especially in the posterior fossa [14]. There is a lack of adequate literature on human studies specifically assessing the long term neurodevelopmental outcomes of birth-related asymptomatic subarachnoid hemorrhage. This is in stark contrast to benign birth-related subdural hemorrhage, where studies have established that there are no significant long-term neurodevelopmental outcomes [12].

Studies have also shown that birth-related subdural hemorrhage measures between 0.5–4.5 mm in width in most cases [5, 13, 15]. A large majority of these are actually less than 3.0 mm in width [2]. Larger hemorrhages are suspicious for traumatic subdural hemorrhage. Our findings were concordant with these studies, with the subdural hemorrhage measuring 3 mm or less in 95.5% of cases. The distribution of birth-related subdural hemorrhage has been well established in literature. These bleeds are most common in the posterior cranium, including posterior fossa, along the tentorium, posterior parieto-occipital regions and along the posterior falx [5, 15, 9]. This distribution is in keeping with the distribution of the fetal dural venous plexus [7]. In our study too, 81.8% of neonates with subdural hemorrhage had bleeds in the posterior fossa. Among the 42.4% of neonates with supratentorial hemorrhages, 96.4% of these were in the parieto-occipital region including the posterior interhemispheric region. This distribution of birth-related subdural hemorrhage is concordant with the existing literature. This posterior cranial distribution of the subdural hemorrhage is useful in differentiating birth-related subdural hemorrhage from subdural hemorrhage due to abusive head trauma. In abusive head trauma, the subdural hemorrhage is often unilaterally or bilaterally hemispheric in distribution, typically located along the bilateral convexities and along the interhemispheric fissure [16]. These are particularly suspicious if they are of mixed signal or density rather than homogeneous [17]. Posterior interhemispheric subdural hemorrhage is seen with birth-related subdural hemorrhage, while subdural hemorrhage associated with abusive head trauma tends to extend into the anterior interhemispheric fissure. However, there are some studies that argue that this feature is not specific and may also be seen in birth-related subdural hemorrhage [17]. We also observed that the incidence of unilateral birth-related subdural hemorrhage in supratentorial compartment was more common on the left side as compared to the right (25% vs 7.1%), although majority of bleeds were bilateral. We believe this may be due to the normal mechanism of delivery favoring specific positions and direction of rotation of the fetal head.

Associated features:

Likewise, there are typically no associated injuries in birth-related subdural hemorrhage such as cortical bridging vein injury, cerebral parenchymal injury or injury to cervical spinal ligaments. These associations are more often seen with abusive head trauma related subdural hemorrhage [16, 18]. Shearing injuries to the gray-white matter junction can cause diffuse axonal injury and cerebral edema [6, 18] in abusive head trauma. Parenchymal injury in birth-related intracranial hemorrhage is limited to minimal parenchymal hemorrhages in the parieto-occipital parenchyma, as seen in 12.1% of our subjects with all hemorrhages measuring 2 mm or less in maximum dimension. The repetitive flexion, extension and rotational movements of the head with or without impact during abusive head trauma in combination with the larger neonatal head-to-body size ratio can result in cervical spinal ligament injuries. In particular, injuries to the posterior spinal ligamentous complex, predominantly the nuchal ligament, posterior atlanto-occipital ligament and posterior atlanto-axial ligament, are frequently identified in abusive head trauma [19]. In our study, we did not find these features in any of the neonates with birth-related intracranial hemorrhage.

Spinal canal subdural hemorrhage:

Another typical differentiating feature between birth-related subdural hemorrhage and abusive head trauma related subdural hemorrhage is the presence of subdural hemorrhage in the spinal canal. This is much more likely to be seen with abusive head trauma related subdural hemorrhage than with birth-related subdural hemorrhage [20]. Some studies show presence of spinal canal subdural hemorrhage in up to 60% of cases of abusive head trauma, with cervical spinal subdural hemorrhage in up to 24% [20]. This feature is also quite specific for abusive head trauma as spinal canal subdural hemorrhage is seen in only around 1% of cases of accidental head trauma [20]. The importance of subdural hemorrhage in the spinal canal in abusive head trauma has been well-recognized in recent years. Recent studies have recommended whole spine MRI rather than cervical spine MRI in all children with suspected abusive head trauma due to the diagnostic importance of recognizing spinal subdural hematomas [21]. With available limited literature on spinal subdural hemorrhages in birth-related intracranial hemorrhage, detectable spinal subdural hemorrhage in association with birth-related subdural hemorrhage appears to be extremely uncommon [20]. This has been reported only in a single case in literature [2]. In our study, there were no cases of spinal subdural hemorrhage in the visualized cervical spinal canal in the cohort. The thoracolumbar spine was not imaged in our study.

Sequential signal evolution for timing of hemorrhages:

All the hemorrhages that we could visualize on MRI were hyperintense on T1W sequence. It is possible that T1 isointense or hypointense subdural hemorrhages were underappreciated. However, hemorrhages usually remain hyperintense on T1W until the late subacute stage (about 4 weeks) and so T1W is very useful for detection of both birth-related and traumatic hemorrhages. We found that changes in T2W signal followed a characteristic pattern in our study. All bleeds were hypointense on T2W until 13 days of age. Bleeds between 13 and 21 days were variable in signal intensity with 26.9% of them showing mixed signal with the rest being T2W hypointense. Beyond 21 days, only a few bleeds were visualized and these were T2W hypointense. On DWI, all bleeds <13 days were hypointense. Beyond this period, some were hypointense while some were hyperintense. Based on these findings, the characteristic evolution of subdural hemorrhage from birth related trauma can be interpreted. The bleed remains hypointense on T2W for the first 2 weeks and does not show diffusion restriction when it is acute. It then becomes progressively mixed in signal from 2–3 weeks in the subacute phase and may show diffusion restriction. In the chronic phase beyond 3 weeks, bleeds are only visible in a small number of neonates and these bleeds do not show restricted diffusion. We believe that this predictable pattern could, in certain cases, be used to identify the timing of a newly detected subdural bleed as being either birth-related or postnatal depending on the signal pattern of the bleed.

When compared to CT, MRI is more sensitive in identifying the presence and chronicity of bleeds [2, 22]. Further, MRI is also more sensitive in detecting spinal canal subdural hemorrhage and other associated features that may point to a diagnosis of abusive head trauma [21]. However, MRI is typically deferred until the condition of the child is stable enough and CT scan may be the only modality feasible in an emergency setting. Typically, in a case of abusive head trauma, the subdural blood tends to be mixed in density on computed tomography (CT) due to admixture of fresh blood with clotted and organizing blood [18]. Sometimes, there may be arachnoid tears which lead to admixture of cerebrospinal fluid with the hemorrhage leading to a subdural hematohygroma [23]. In birth-related subdural hemorrhage, the bleeds are usually relatively homogeneous in appearance in the early stages [5].

One major challenge with using MRI for timing the subdural hemorrhages is the fact that most studies on predictable pattern of signal changes in hemorrhage have been performed on parenchymal hemorrhages. Studies have shown that subdural hemorrhages may not necessarily follow this trajectory [22]. This was also observed in our study, in the sense that a number of cases showed T2W hypointense bleeds even beyond the third week. Various theories that have been postulated to explain this and the actual reason may be a combination of various factors. Presence of a higher oxygen tension within subdural hemorrhages may delay the bleed evolution [24]. Clotting mechanisms and development of neomembranes may also influence the signal. Further complicating this are associated dural injures such as arachnoid tears and avulsion of arachnoid granulations that lead to admixture of cerebrospinal fluid [23, 25]. With abusive head trauma, there is also a tendency for recurrent rebleeding into the hemorrhage [6, 26]. Birth-related subdural hemorrhage may be relatively more predictable than abusive head trauma related subdural hemorrhage due to the small size of the hemorrhages and absence of arachnoid injuries [1, 5]. Observations from our own study show that such timing of the hemorrhages is not entirely reliable and would be useful only if used in conjunction with other associated features in order to differentiate birth-related subdural hemorrhage from traumatic subdural hemorrhage. Further studies are necessary in order to assess the accuracy of neuroradiologic timing of birth-related subdural hemorrhages.

Duration:

Subdural hemorrhages in birth-related intracranial hemorrhage were observed in neonates up to 25 days of age in our study. There were no detectable hemorrhages beyond 25 days of age, although there were subjects up to 54 days of age in our study. This potentially may suggest that the majority of birth-related subdural hemorrhages resolves sufficiently in the first 4 weeks of life to be undetectable on imaging after the first month of postnatal life. This timeline often holds good because these bleeds are small in size and resolve completely. Subsequently, any subdural hemorrhages, if detected on imaging beyond 4 weeks of age, would need to be evaluated for traumatic subdural hemorrhage from accidental or non-accidental trauma even if they are small and are distributed in those locations typical for birth-related subdural hemorrhage. Other studies have shown similar findings [27, 28], with almost all birth-related subdural hemorrhage resolving in the first month of life.

Limitations:

Our study had a number of limitations. A major drawback was that imaging was performed, at a minimum, 9 days after birth. This may underestimate the incidence of birth-related hemorrhages, since smaller birth related hemorrhages may have become undetectable by the time of imaging. Cortical vein thrombosis which may have recanalized by this time could have also been missed. While the infants included in this study were clinically asymptomatic, other causes of pathology leading to intracranial hemorrhages were not excluded. The asymptomatic subdural hemorrhages were assumed to have occurred at birth for the purpose of interpreting the sequential signal evolution. However, asymptomatic subdural hemorrhages occurring before or after birth, while uncommon, cannot be ruled out. Although we evaluated for cervical spinal subdural hemorrhage, the rest of the spine was not imaged. Subdural hemorrhages in the lower thoracic and lumbar spine, which are more common than cervical spinal subdural hemorrhage, may have been missed. In addition, Short Tau Inversion Recovery (STIR) or T2W fat-suppressed images of the cervical spine were not acquired and subtle cervical spinal ligamentous injury may have been missed on T1W and T2W sequences. Only T1W, T2W and diffusion sequences of the brain were performed. Gradient or susceptibility sequences were not performed. Subtle hemorrhages and thrombosed cortical bridging veins might have been missed, which could have been detected with gradient or susceptibility imaging. Fluid Attenuated Inversion Recovery (FLAIR) sequence was also not performed, which limited the sensitivity for detecting hematohygromas. Although our study had a larger sample size than most prior studies, it was still relatively small. We did not sequentially image the bleeds in order to track their evolution. Lack of follow-up studies to sequentially follow the changes in signal intensity of the bleeds over time significantly limits the interpretation of sequential evolution of signal characteristics. In cases where no bleeds were detected at the time of imaging, we assumed that there were no bleeds. However, this could not differentiate cases with resolved bleeds from cases in which bleeds were never present.

Conclusion:

Birth-related subdural hemorrhage is small in size in most cases and is a common and benign entity that occurs in over a third of newborn infants. They have a characteristic distribution, being predominantly in the posterior fossa and in the posterior cranium if supratentorial. Within the limitations of the study, no associated cervical spinal subdural hemorrhages, cervical spinal ligamentous injury or cortical bridging vein injury were observed. The absence of these findings is reassuring, as their presence would raise concern for traumatic subdural hemorrhages, particularly those due to abusive head trauma in the neonatal age group. Neonates born by vaginal delivery, in particular assisted vaginal delivery, have a significantly higher incidence of birth-related subdural hemorrhage as compared to cesarean section. Although not completely reliable, birth-related subdural hemorrhages tend to follow a characteristic pattern of signal changes on MRI. No case of birth related subdural hemorrhage was identified beyond 25 days of life on MRI brain studies, implying complete resolution within the first month of life.

Data Availability:

The data that support the findings of this study are not openly available due to reasons of sensitivity and are available from the corresponding author upon reasonable request. Data are located in controlled access data storage at Arkansas Children’s Hospital.