Optical Detection of Intracranial Pressure and Perfusion Changes in Neonates With Hydrocephalus

Abstract

Objective:

To demonstrate that a novel non-invasive index of intracranial pressure (ICP) derived from diffuse optics-based techniques is associated with intracranial hypertension.

Study design:

We compared non-invasive and invasive ICP measurements in infants with hydrocephalus. Infants born term and preterm were eligible for inclusion if clinically determined to require cerebrospinal fluid (CSF) diversion. Ventricular size was assessed preoperatively via ultrasound measurement of the fronto-occipital (FOR) and fronto-temporal (FTHR) horn ratios. Invasive ICP was obtained at the time of surgical intervention with a manometer. Intracranial hypertension was defined as invasive ICP ≥15 mmHg. Diffuse optical measurements of cerebral perfusion, oxygen extraction, and non-invasive ICP were performed preoperatively, intraoperatively, and postoperatively. Optical and ultrasound measures were compared with invasive ICP measurements, and their change in values after CSF diversion were obtained.

Results:

We included 39 infants; 23 had intracranial hypertension. No group difference in ventricular size was found by FOR (p=0.93) or FTHR (p=0.76). Infants with intracranial hypertension had significantly higher non-invasive ICP (p=0.02) and oxygen extraction fraction (p=0.01) compared with infants without intracranial hypertension. Increased cerebral blood flow (p=0.005) and improved oxygen extraction fraction (P < .001) after CSF diversion were only observed in infants with intracranial hypertension.

Conclusions:

Non-invasive diffuse optical measures (including a non-invasive ICP index) were associated with intracranial hypertension. The findings suggest impaired perfusion from intracranial hypertension was independent of ventricular size. Hemodynamic evidence of the benefits of CSF diversion was seen in infants with intracranial hypertension. Non-invasive optical techniques hold promise for aiding the assessment of CSF diversion timing.

Hydrocephalus is the most common neuropathology treated in pediatric neurosurgery¹, and is commonly associated with impaired cognitive outcomes.^2–5 Among the causes of these deficits are periventricular white matter insult and hypoxic-ischemic injury to brain parenchyma that arise from intracranial hypertension.^{4, 6, 7} Therefore, the timing of neurosurgical intervention for CSF diversion to relieve intracranial hypertension is important. This timing, however, is currently highly variable among pediatric neurosurgeons^{7, 8} and bedside non-invasive detection of intracranial hypertension could aid clinical decision-making.

Numerous techniques have been proposed for this end^9–12 and although promising, they are not yet reliable enough for clinical use. Recent proof-of-concept studies have further utilized machine learning to model the relationship between optical measurements of cerebral blood flow waveform morphology and intracranial pressure (ICP).^{13, 14} In this manuscript, we present a pilot study of non-invasive optical methods for detection of intracranial hypertension in neonatal hydrocephalus.

We combined two optical techniques, frequency-domain diffuse optical spectroscopy (FD-DOS) and diffuse correlation spectroscopy (DCS), to derive non-invasive indices of cerebral blood flow (CBF), oxygen extraction fraction (OEF), and ICP.^15–19 We assessed these methods in infants receiving CSF diversion. We hypothesized that: (a) FD-DOS/DCS measurements can detect intracranial hypertension in neonatal hydrocephalus; (b) intracranial hypertension is associated with impaired cerebral oxygen delivery (ie, diminished CBF and elevated OEF as measured by FD-DOS/DCS); and (c) intracranial hypertension is associated with a favorable metabolic response to CSF diversion (i.e., increased CBF and decreased OEF). Ultimately, successful implementation of bedside FD-DOS/DCS monitoring of ICP may facilitate individualized patient-specific management of infant hydrocephalus.

Methods:

The Children’s Hospital of Philadelphia (CHOP) institutional review board approved this study, and written informed consent was obtained from the guardians of all participants. We screened all patients admitted to the neonatal intensive care unit (NICU) between 9/1/2018 to 3/15/2020. Patients were eligible for inclusion if: (1) their age was <2 months; (2) they had a diagnosis of hydrocephalus; and (3) they met local surgical criteria^{20, 21} for CSF diversion to manage hydrocephalus. Infants were excluded if an invasive ICP measurement was not performed. CSF diversion and shunt timing were based on weight, brain ultrasound measurements of ventriculomegaly, and clinical exam criteria.^{20, 21} Demographic information including sex, weight and age at surgery, race, diagnosis, and head circumference were abstracted from the clinical chart.

Enrolled subjects received a preoperative FD-DOS/DCS measurement over the frontal lobe (Figure 1; available at www.jpeds.com) within seven days of surgery. We additionally examined the closest preoperative brain ultrasound to surgery; the timing of this ultrasound was based on routine clinical care, but to be eligible for use in the study, it had to have been obtained within ten days of surgery. At the time of surgery, after induction of anesthesia, an intraoperative optical measurement was obtained. During surgical CSF diversion, a brain cannula was used to make the tract to the ventricle, and the attending neurosurgeon immediately acquired an invasive ICP measurement with a manometer attached to the end of the cannula (zero-point level was the foramen of Monro). A post-operative optical measurement was obtained within seven days of surgery. Accompanying each optical measurement, a single blood pressure was acquired with a clinical, oscillometric cuff. Calculation of CBF, cerebral oxygen saturation (StO₂), and total hemoglobin concentration (THC) was performed.¹⁶ The optical instrument and methods are further described in the Appendix (available at www.jpeds.com).

Figure 1.

(A) Schematic of optical probe placement, as it was manually held against the forehead during data acquisition. (B) Timeline of the study’s preoperative, intraoperative, and postoperative measurements encompassing cerebrospinal fluid (CSF) diversion surgery. (C) Study design for intraoperative measurements, which include right and left hemisphere optical measurements, a cuff blood pressure (BP) measurement, and an invasive intracranial pressure (ICP) measurement obtained during cerebrospinal fluid (CSF) diversion surgery. (D) Study design for preoperative and postoperative measurements.

Our primary study measurement parameter was a non-invasive ICP index (nICP) calculated from FD-DOS/DCS. Briefly, ICP was estimated based on the relationship between microvascular CBF and blood pressure (i.e., ICP is the extrapolated arterial blood pressure at which CBF reaches zero). This value, which is also known as the critical closing pressure, can be calculated using the pulsatility indices (PI) of both CBF and blood pressure (PI_CBF and PI_BP).¹⁷ The pulsatility index of a measurement is its value in systole minus its value in diastole divided by its mean. Then, as described elsewhere,¹⁷ we defined: nICP = 0.6 · MAP(1 – PI_BP/PI_CBF). Here, MAP is the mean arterial blood pressure. MAP and PI_BP were calculated from the oscillometric arterial blood pressure measurement, and PI_CBF was calculated from the DCS data using Fourier analysis.¹⁷

First, we compared intraoperatively measured nICP to invasively measured ICP (the gold standard); the optical measurement was made several minutes before the invasive ICP measurement under similar anesthetic conditions (Figure 1, B). We hypothesized that nICP would be significantly correlated with invasively-measured ICP and that elevated nICP would be associated with intracranial hypertension; intracranial hypertension was defined as an invasive ICP measurement of ≥ 15 mm Hg. This threshold for dichotomization is consistent with the current literature¹¹ and was chosen based on extrapolation of adult traumatic brain injury data.

We also investigated whether other optical biomarkers, obtained intraoperatively, were associated with intracranial hypertension. Namely, we measured StO₂, THC, and CBF, and we derived oxygen extraction fraction (OEF) and the cerebral metabolic rate of oxygen consumption (CMRO₂) using standard calculations^{16, 22} (Appendix). We compared invasively measured ICP with brain ultrasound measurements of the fronto-occipital horn ratio (FOR) and fronto-temporal horn ratio (FTHR).^{23, 24} FOR and FTHR are standard metrics of ventricular size; they were both measured post hoc by an experienced neuroradiologist blinded to the clinical status of the patient. Finally, we investigated whether the average rate of change of head circumference over the week prior to surgery was associated with intracranial hypertension.

Statistical Analyses

For the main analyses, mean optical measures across both hemispheres were used. Statistical calculations were performed with MATLAB R2018a (Mathworks, Inc, Natick, MA, USA). All statistical tests were two-sided, and p-values <0.05 were used to deem significance.

We assessed whether infants with and without intracranial hypertension differed in intraoperative optical biomarkers (nICP, StO₂, THC, CBF, OEF, and CMRO₂) or ultrasound biomarkers (FOR and FTHR, measured preoperatively) using t-tests. We also used linear regression and Bland-Altman analyses to investigate agreement between intraoperative nICP and invasive ICP. Finally, we used linear regression analyses to estimate the associations between preoperative ultrasound metrics (FOR and FTHR) and invasive ICP.

For our secondary analysis, we examined pre- to post-operative changes in optically measured nICP, StO₂, THC, CBF, OEF, and CMRO₂. Of note, because the intraoperative optical measurement occurred immediately prior to CSF diversion, we used the measured intraoperative to postoperative changes for this analysis to best isolate hemodynamic effects associated with CSF diversion. Specifically, the relative postoperative/intraoperative ratios were computed for OEF, CBF, THC, and CMRO₂, and the relative postoperative-to-intraoperative differences were computed for nICP and StO₂. Paired t-tests were employed to determine whether mean changes in these variables were different from zero, for both infants with intracranial hypertension and infants without intracranial hypertension. Last, we assessed agreement between intraoperative and preoperative FD-DOS/DCS measurements with intraclass correlation coefficients (Appendix).²⁵

Results:

During the study period, 52 patients were eligible for enrollment (Figure 2; available at www.jpeds.com). Of these, 5 parents refused consent, 1 was not approached for consent due to staff availability, and 7 patients did not have invasive ICP measurements performed. Thus, 39 patients were included in the analyses. There were 26 deviations from the intended study protocol requiring some patients to be excluded from some analyses. Namely, 5 patients did not have a preoperative optical measurement, 11 did not have an intraoperative optical measurement, and 10 did not have a postoperative measurement. The reasons for missed measurements were FD-DOS/DCS machine availability (1 preoperative; 4 intraoperative; 2 postoperative), staff availability (4 preoperative; 7 intraoperative; 3 postoperative), NICU discharge (4 postoperative), and clinical instability (1 postoperative). In 9 patients, a brain ultrasound was not performed within 10 days of the operation. The preoperative optical measurement occurred a median of 2.5 days (IQR: 2–4 days) before surgery. Similarly, the postoperative optical measurement occurred a median of 2.5 days (IQR: 2–4 days) after surgery. Finally, the brain ultrasound scans occurred a median of 4.5 days (IQR: 3–8 days) prior to surgery.

Figure 2.

Patient flow chart.

There was a wide variation in invasively-measured ICP at the time of CSF diversion from 5 to 30 mmHg; 23 patients (59%) met the study definition of intracranial hypertension (Table I). Compared with infants without intracranial hypertension, a higher proportion of infants with intracranial hypertension received the ventriculoperitoneal shunting method of CSF diversion (p = 0.03). The use of general anesthesia during CSF diversion also differed between groups (p = 0.02). Infant weight at the time of CSF diversion was lower in the non-intracranial hypertension group, but this difference was not significant (p = 0.06). Cerebral perfusion pressure at the time of surgery was lower in the intracranial hypertension group (p = 0.04). No other differences were identified between groups.

Primary Outcome: Association of Noninvasive Biomarkers with Intracranial Hypertension

The primary outcome analysis involved 28 patients with both intraoperative FD-DOS/DCS and invasive ICP data. Of these patients, 18 (64%) had intracranial hypertension. Non-invasive ICP was significantly higher in the infants with intracranial hypertension (p = 0.02, Figure 3, A). Patients with intracranial hypertension also had significantly lower StO₂ (p = 0.01, Figure 3, B) and higher OEF (p = 0.01, Figure 3C). THC and CBF were lower in the intracranial hypertension group, although these differences were not significant (Figure 3, D and E). No difference in CMRO₂ was observed between groups (p = 0.56, Figure 3, F).

Figure 3.

Association of noninvasive optical and ultrasound biomarkers with intracranial hypertension defined as an invasive intracranial pressure (ICP) measurement of ≥15 mmHg. Units for CBF and CMRO₂ indices are 10⁻⁹ cm²/s and 10⁻⁴ μM cm²/s, respectively. All boxplots show the median ± interquartile range.

Thirty patients had invasive ICP data and preoperative FOR and FTHR measurements for analysis (20 of these patients were in the primary outcome analysis). Of these patients, 16 (53%) had intracranial hypertension. In contrast to optical biomarkers, there were no differences in preoperative ultrasound metrics (median FOR: p=0.93 and FTHR: p=0.76) between the groups with and without intracranial hypertension (Figure 3, G and H).

Secondary Analyses

Linear regression analysis showed nICP to be significantly correlated with ICP (p = 0.01, Figure 4, A), although the correlation was moderate (R = 0.48), and the slope of the line of best fit between nICP and ICP was 0.50 (95% CI: 0.13 to 0.87). In Bland-Altman analysis, the mean difference of nICP and ICP was 1.0 mmHg (95% CI: −10.5 to 12.4), not significantly different from zero (p = 0.39). Conversely, increased FOR and FTHR measured by brain ultrasound were not associated with increasing invasively-measured ICP (FOR: slope −0.0005 (95% CI: −0.0075 to 0.0065), R = −0.03, p = 0.89 and FTHR: slope −0.0001 (95% CI: −0.0056 to 0.0054), R = −0.01, p = 0.97; Figure 4, C and D). Although head circumference was not associated with intracranial hypertension (Table 1), it did modestly increase between the time of ultrasound and surgery by a median of 0.9 cm (IQR: 0.1, 1.5).

Figure 4.

(A) Intraoperative non-invasive ICP (nICP) was correlated with ICP acquired during CSF diversion surgery (solid line is the linear best fit, dashed line is line of unity). (B) Bland-Altman plot of the difference between nICP and ICP (solid line indicates the mean difference; dashed lines indicate the 95% limits of agreement, i.e., the mean ± 1.96 times the standard deviation of the difference). (C) Preoperative fronto-occipital horn ratio (FOR) and (D) fronto-temporal horn ratio (FTHR) as measured on brain ultrasound were not correlated with ICP (solid lines are the linear best fit).

Table 1.

Demographics of patients with and without intracranial hypertension

	No Intracranial Hypertension (ICP < 15 mmHg)	Intracranial Hypertension (ICP ≥ 15 mmHg)	p-value
N	16	23
Female sex	7 (44%)	8 (35%)	0.74
Gestational age (weeks)	32.4 (26.5, 35.9)	32.6 (24.9, 37.1)	0.89
Weight at time of CSF diversion surgery (grams)	2525 (2130, 2900)	3080 (2358, 3600)	0.06
Race
White	9 (56%)	15 (65%)	0.74
Black	7 (44%)	8 (35%)	0.74
Born preterm	13 (81%)	16 (70%)	0.48
Age at time of CSF diversion surgery (days)
Preterm infants	27 (22, 25)	64 (29, 112)	0.12
Term infants	5 (2, 8)	6 (5, 10)	0.72
Etiology of hydrocephalus
Post-hemorrhagic	8 (50%)	8 (35%)	0.51
Congenital	1 (6%)	4 (17%)	0.63
Myelomeningocele (postnatal repair)	4 (25%)	8 (35%)	0.73
Myelomeningocele (prenatal repair)	1 (6%)	2 (9%)	>0.99
Other	2 (13%)	1 (4%)	0.56
CSF diversion Type
Ventriculoperitoneal shunt	11 (69%)	22 (96%)	0.03
Ommaya reservoir	4 (25%)	0 (0%)	0.02
External ventricular drain	0 (0%)	1 (4%)	>0.99
Endoscopic third ventriculostomy	1 (6%)	0 (0%)	0.41
Inhalational volatile anesthetic used for CSF diversion surgery	12 (75%)	23 (100%)	0.02
Received serial drainage prior to CSF diversion surgery	2 (13%)	5 (22%)	0.68
Intraoperative head circumference percentile	77 (24, 94)	92 (46, 97)	0.44
Rate of change of head circumference over week prior to CSF diversion surgery (cm/day)	0.17 (0.02, 0.24)	0.21 (0.10, 0.28)	0.33
Intraoperative MAP (mmHg)	40 (34, 42)	44 (38, 46)	0.22
Intraoperative cerebral perfusion pressure (mmHg)	28 (23, 31)	24 (17, 28)	0.04

Data are shown as number (percent) or median (interquartile range), as appropriate (p-values were computed via the Fisher exact test and the Wilcoxon rank sum test, as appropriate). Intraoperative mean arterial pressure (MAP) is the mean cuff blood pressure between anesthesia induction and surgical incision. Intraoperative cerebral perfusion pressure is the difference between MAP and ICP.

Eighteen patients had both intraoperative and postoperative FD-DOS/DCS data for analysis of changes in optical biomarkers; 10 had intracranial hypertension (56%). Patients who had intracranial hypertension demonstrated postoperative decreases in nICP (p = 0.02) and OEF (p < 0.001) and postoperative increases in CBF (p = 0.005), CMRO₂ (p = 0.02), THC (p = 0.003), and StO₂ (p < 0.001). In contrast, patients without intracranial hypertension showed smaller physiologic effects of surgery with no statistically significant changes in FD-DOS/DCS measures; CMRO₂ increased after surgery in these patients (p = 0.05, Figure 5). Finally, though MAP was not associated with intracranial hypertension (Table 1), the median intraoperative MAP of 41 mmHg (IQR: 37, 45) was lower than both the median postoperative MAP of 55 mmHg (IQR: 49, 63) and the median preoperative MAP of 48 mmHg (IQR: 43, 60).

Figure 5.

Changes in optical biomarkers from before to after cerebral spinal fluid (CSF) diversion surgery, dichotomized by our definition of intracranial hypertension as an intracranial pressure (ICP) of ≥ 15 mmHg. ΔnICP is the difference between postoperatively and intraoperatively measured non-invasive ICP (nICP). Relative oxygen extraction fraction (rOEF), cerebral blood flow (rCBF), cerebral metabolic rate of oxygen (rCMRO₂), and total hemoglobin (rTHC) are expressed as postoperative-to-intraoperative ratios. ΔStO₂ is the difference between postoperatively and intraoperatively measured tissue oxygen saturation. P-values indicate whether the mean change was different from zero. All boxplots show the median ± interquartile range.

Repeatability of the optical measures was assessed with intraclass correlation coefficients. For all optical biomarkers, good agreement was noted between measurements in each hemisphere and between preoperative and intraoperative measures (Table 2; available at www.jpeds.com).

Table 2.

Intraclass correlation coefficients (ICCs) for optical biomarkers within and across measurements. ICCs for non-invasive ICP (nICP), tissue oxygen saturation (StO2), total hemoglobin concentration (THC), cerebral blood flow (CBF), oxygen extraction fraction (OEF), and the cerebral metabolic rate of oxygen (CMRO2) were all good-to-excellent.

	Left & Right hemisphere (n=90)	Pre- & Intra-operative (n = 23)
nICP	0.86 (0.80, 0.91)	0.72 (0.44, 0.87)
StO 2	0.74 (0.62, 0.82)	0.80 (0.56, 0.91)
THC	0.88 (0.82, 0.92)	0.89 (0.73, 0.95)
CBF	0.87 (0.80, 0.91)	0.84 (0.59, 0.94)
OEF	0.74 (0.62, 0.82)	0.80 (0.57, 0.92)
CMRO 2	0.72 (0.60, 0.81)	0.81 (0.55, 0.92)

Data are presented as the ICC and 95% confidence interval.

Discussion:

In this pilot study, we demonstrated feasibility of non-invasive, bedside FD-DOS/DCS measurements in infants with hydrocephalus. Moreover, we have shown that these measurements can provide important information not attainable with current clinical proxies for elevated ICP. Notably, our non-invasive ICP index, nICP, was significantly associated with intracranial hypertension, whereas the current clinical standards of fronto-occipital horn ratio (FOR) and fronto-temporal horn ratio (FTHR), calculated from brain ultrasound, were not associated with intracranial hypertension. These results suggest that ventricular size alone is inadequate as a predictor of intracranial hypertension in neonatal hydrocephalus.

In the present investigation, we found that intracranial hypertension was associated with evidence of impaired cerebral oxygen delivery; OEF was higher in patients with intracranial hypertension. A higher OEF indicates that more oxygen is extracted from the delivered blood. Direct evidence of decreased oxygen delivery was not found (i.e., CBF and THC were not significantly lower in patients with intracranial hypertension). However, both CBF and THC showed a trend towards a statistically significant decrease, with p-values affected by a single patient with intracranial hypertension with a high CBF and THC. Although one might expect that increases in MAP can compensate for intracranial hypertension to maintain cerebral perfusion pressure, cerebral perfusion pressure was lower in the group with intracranial hypertension. In an extreme case, decreased oxygen delivery would result in ischemia and lower oxygen metabolism. On average, decreased oxygen delivery was compensated by increased oxygen extraction (i.e., elevated OEF); consequently, there was no measured difference in intraoperative CMRO₂ index between the groups with and without intracranial hypertension. However, this overall assessment may miss individual patients with critically diminished oxygen delivery.

FD-DOS/DCS measurements also supported our hypothesis that CSF diversion in patients with intracranial hypertension was associated with improved biomarkers of cerebrovascular function. CSF diversion resulted in a decrease in nICP, as would be expected for a successful surgical result. These patients experienced an increase in CBF, an increase in blood volume, and a decrease in OEF. CSF diversion thus resulted in improved oxygen delivery and a return towards more normative metabolism.

Our finding that patients without intracranial hypertension did not show a physiologic response to CSF diversion is interesting and has implications for clinical care. Patients with ICP <15 mmHg had no change in nICP, CBF or other measures of oxygen delivery after CSF diversion. When selecting patients for invasive surgery to relieve intracranial hypertension, we desire tools that are both sensitive (to avoid delaying surgery beyond where increased ICP impairs oxygen delivery) and specific (to avoid surgery in patients where conservative management may be appropriate). The variance in ICP observed at the invasive management time-point indicates that there is significant variability in current clinical practice. Future work with longitudinal FD-DOS/DCS measurements holds potential to provide a more complete picture of neonatal hydrocephalus pathophysiology, thereby enabling improved prognostication and determination of patient-specific risks and benefits of surgical intervention.

Our results expand on prior studies of optical neuromonitoring in this population, which commonly relied on detection of diminished cerebral oxygenation using near-infrared spectroscopy (NIRS).^26–30 Whereas cerebral oxygenation has been associated with ventricular dilation³⁰ and has been demonstrated to improve after CSF diversion,^26–29 NIRS measurements have limited reproducibility and reliability.^{31, 32} FD-DOS uses a more quantitative measurement of StO₂ that accounts for tissue scattering.¹⁵ The success in this approach is demonstrated by the high inter-measurement reliability.

Our study had several limitations. In a significant number of patients, optical measurements were not available at all time-points. Similarly, due to variation in clinical care (outside of the research protocol), brain ultrasounds were not always available for analysis. Thus, the sample size was below that initially anticipated, and the various analyses were performed on slightly different populations. Although certain analyses may be under-powered, we do not expect that these deviations resulted in fundamental changes in interpretation. We note that there was a near-significant difference in weight at the time of surgery between the groups with and without intracranial hypertension, which is likely why more Ommaya reservoirs were used in the group without intracranial hypertension. Differences in physiologic responses between Ommaya reservoirs and ventriculoperitoneal shunts need to be investigated in future work. Larger studies in the future are also needed to facilitate multivariate modelling for the prediction of intracranial hypertension based on FD-DOS/DCS measurements and other covariates such as etiology of hydrocephalus, age, head circumference percentile, rate of change of head circumference, apneic spells, bradycardia episodes, and other clinical markers.

We were limited in the timing of ultrasound measures based on routine clinical care, and we did not assess for longitudinal changes in ultrasound metrics. It is possible that preoperative FOR and FTHR measurements obtained immediately prior to surgery would be more predictive of ICP. However, the minor changes observed in head circumference prior to surgery suggest that interval changes in ultrasound also would be small. Thus, although day-of-surgery ultrasounds would likely have slightly higher FOR and FTHR, this would likely not impact the lack of association between ultrasound and ICP that we observed.

Our measurement of nICP has methodological limitations. We relied on a single non-invasive cuff blood pressure measurement to determine systolic, diastolic, and mean arterial pressure. Calculation of nICP would be improved by continuous arterial blood pressure monitoring; however, because invasive blood pressure monitoring is not standard of care at our institution, such data was unavailable to us. In the future, the use of continuous non-invasive blood pressure measurements³³ has the potential to substantially improve the accuracy of the nICP measurement. We note that the use of the average across cuff blood pressure readings from anesthesia induction to incision to calculate nICP did not improve accuracy (the Pearson correlation with invasive ICP decreased from 0.48 to 0.44). There is also scope to improve nICP measurement accuracy through the development of improved tissue models that account for effects of superficial tissue contamination in the FD-DOS/DCS optical signals (which are likely more pronounced in older patients).^{34, 35} A third direction for future research is the use of machine learning strategies.^{13, 14}

Anesthesia is a potential confound for our secondary analysis of intraoperative to postoperative changes. We did observe that patients with intracranial hypertension were more likely to receive general anesthesia with an inhalational volatile anesthetic than patients without intracranial hypertension. Inhalational volatile anesthetics can potentially alter intraoperative cerebral hemodynamics, for example, by decreasing cerebrovascular resistance. Our observed agreement between preoperative and intraoperative CBF measurements (Table 2), however, suggests that the influence of general anesthesia on cerebrovascular resistance is modest. Further, the higher postoperative MAP observed when a patient is awake would most likely elevate ICP and would not impact our observation of decreased nICP after CSF diversion in the intracranial hypertension group.

Finally, we defined intracranial hypertension as an invasive ICP ≥15 mmHg. We arrived at this definition based on prior results in adults and children.^{11, 36–38} However, we acknowledge that no standard definition of intracranial hypertension exists. The gold standard would be a reliable way to predict long-term neurodevelopmental outcomes based on data available pre- or intra-operatively. In the absence of such a gold standard, we believe our dichotomization of patients using 15 mmHg was a reasonable definition. We aspire to conduct larger cohort studies to compare optical biomarkers to long-term neurodevelopmental testing.

We have shown that a novel non-invasive diffuse optical index of ICP was significantly associated with intracranial hypertension in neonatal hydrocephalus. By contrast, ventricular size near the time of CSF diversion surgery was not associated with ICP. Further, the diffuse optical measurements illuminated the effect of intracranial hypertension on cerebral metabolism; the measurements demonstrated that diminished cerebral oxygen delivery and a positive response to CSF diversion were only observed in infants with elevated ICP. Our findings suggest that the use of noninvasive optical biomarkers may potentially improve on current clinical practice by offering better ways to classify patients such that the risks and benefits of surgery can be individualized.

Abbreviations and Acronyms:

BP

blood pressure

CBF

cerebral blood flow

CMRO₂

cerebral metabolic rate of oxygen consumption

CSF

cerebrospinal fluid

DCS

diffuse correlation spectroscopy

FD-DOS

frequency-domain diffuse optical spectroscopy

FOR

fronto-occipital horn ratio

FTHR

fronto-temporal horn ratio

ICP

intracranial pressure

MAP

mean arterial pressure

nICP

non-invasive ICP index

OEF

oxygen extraction fraction

PI

pulsatility index

StO₂

tissue oxygen saturation

THC

total hemoglobin concentration