Reduced complexity of intracranial pressure observed in short time series of intracranial hypertension following traumatic brain injury in adults

Random forest-based prediction of intracranial hypertension in patients with traumatic brain injury

BACKGROUND

Treatment and prevention of intracranial hypertension (IH) to minimize secondary brain injury are central to the neurocritical care management of traumatic brain injury (TBI). Predicting the onset of IH in advance allows for a more aggressive prophylactic treatment. This study aimed to develop random forest (RF) models for predicting IH events in TBI patients.

METHODS

We analyzed prospectively collected data from patients admitted to the intensive care unit with invasive intracranial pressure (ICP) monitoring. Patients with persistent ICP > 22 mmHg in the early postoperative period (first 6 h) were excluded to focus on IH events that had not yet occurred. ICP-related data from the initial 6 h were used to extract linear (ICP, cerebral perfusion pressure, pressure reactivity index, and cerebrospinal fluid compensatory reserve index) and nonlinear features (complexity of ICP and cerebral perfusion pressure). IH was defined as ICP > 22 mmHg for > 5 min, and severe IH (SIH) as ICP > 22 mmHg for > 1 h during the subsequent ICP monitoring period. RF models were then developed using baseline characteristics (age, sex, and initial Glasgow Coma Scale score) along with linear and nonlinear features. Fivefold cross-validation was performed to avoid overfitting.

RESULTS

The study included 69 patients. Forty-three patients (62.3%) experienced an IH event, of whom 30 (43%) progressed to SIH. The median time to IH events was 9.83 h, and to SIH events, it was 11.22 h. The RF model showed acceptable performance in predicting IH with an area under the curve (AUC) of 0.76 and excellent performance in predicting SIH (AUC = 0.84). Cross-validation analysis confirmed the stability of the results.

CONCLUSIONS

The presented RF model can forecast subsequent IH events, particularly severe ones, in TBI patients using ICP data from the early postoperative period. It provides researchers and clinicians with a potentially predictive pathway and framework that could help triage patients requiring more intensive neurological treatment at an early stage.

Predicting the Severity and Discharge Prognosis of Traumatic Brain Injury Based on Intracranial Pressure Data Using Machine Learning Algorithms

OBJECTIVE

This study aimed to explore the potential of employing machine learning algorithms based on intracranial pressure (ICP), ICP-derived parameters, and their complexity to predict the severity and short-term prognosis of traumatic brain injury (TBI).

METHODS

A single-center prospectively collected cohort of neurosurgical intensive care unit admissions was analyzed. We extracted ICP-related data within the first 6 hours and processed them using complex algorithms. To indicate TBI severity and short-term prognosis, Glasgow Coma Scale score on the first postoperative day and Glasgow Outcome Scale-Extended score at discharge were used as binary outcome variables. A univariate logistic regression model was developed to predict TBI severity using only mean ICP values. Subsequently, 3 multivariate Random Forest (RF) models were constructed using different combinations of mean and complexity metrics of ICP-related data. To avoid overfitting, five-fold cross-validations were performed. Finally, the best-performing multivariate RF model was used to predict patients' discharge Glasgow Outcome Scale-Extended score.

RESULTS

The logistic regression model exhibited limited predictive ability with an area under the curve (AUC) of 0.558. Among multivariate models, the RF model, combining the mean and complexity metrics of ICP-related data, achieved the most robust ability with an AUC of 0.815. Finally, in terms of predicting discharge Glasgow Outcome Scale-Extended score, this model had a consistent performance with an AUC of 0.822. Cross-validation analysis confirmed the performance.

CONCLUSIONS

This study demonstrates the clinical utility of the RF model, which integrates the mean and complexity metrics of ICP data, in accurately predicting the TBI severity and short-term prognosis.

Central autonomic network and early prognosis in patients with disorders of consciousness

The central autonomic network (CAN) plays a crucial role in modulating the autonomic nervous system. Heart rate variability (HRV) is a valuable marker for assessing CAN function in disorders of consciousness (DOC) patients. We used HRV analysis for early prognosis in 58 DOC patients enrolled within ten days of hospitalization. They underwent a five-minute electrocardiogram during baseline and acoustic/visual stimulation. The coma recovery scale-revised (CRS-R) was used to define the patient's consciousness level and categorize the good/bad outcome at three months. The high-frequency Power Spectrum Density and the standard deviation of normal-to-normal peaks in baseline, the sample entropy during the stimulation, and the time from injury features were used in the support vector machine analysis (SVM) for outcome prediction. The SVM predicted the patients' outcome with an accuracy of 96% in the training test and 100% in the validation test, underscoring its potential to provide crucial clinical information about prognosis.

Fractal Analysis of the Cerebrovascular System Pathophysiology

The cerebrovascular system is characterized by parameters such as arterial blood pressure (ABP), cerebral perfusion pressure (CPP), and cerebral blood flow velocity (CBFV). These are regulated by interconnected feedback loops resulting in a fluctuating and complex time course. They exhibit fractal characteristics such as (statistical) self-similarity and scale invariance which could be quantified by fractal measures. These include the coefficient of variation, the Hurst coefficient H, or the spectral exponent α in the time domain, as well as the spectral index ß in the frequency domain. Prior to quantification, the time series has to be classified as either stationary or nonstationary, which determines the appropriate fractal analysis and measure for a given signal class. CBFV was characterized as a nonstationary (fractal Brownian motion) signal with spectral index ß between 2.0 and 2.3. In the high-frequency range (>0.15 Hz), CBFV variability is mainly determined by the periodic ABP variability induced by heartbeat and respiration. However, most of the spectral power of CBFV is contained in the low-frequency range (<0.15 Hz), where cerebral autoregulation acts as a low-pass filter and where the fractal properties are found. Cerebral vasospasm, which is a complication of subarachnoid hemorrhage (SAH), is associated with an increase in ß denoting a less complex time course. A reduced fractal dimension of the retinal microvasculature has been observed in neurodegenerative disease and in stroke. According to the decomplexification theory of illness, such a diminished complexity could be explained by a restriction or even dropout of feedback loops caused by disease.

Intracranial Pressure Variability: A New Potential Metric of Cerebral Ischemia and Energy Metabolic Dysfunction in Aneurysmal Subarachnoid Hemorrhage?

BACKGROUND

It was recently reported that lower intracranial pressure variability (ICPV) is associated with delayed ischemic neurological deficits and unfavorable outcomes in patients with aneurysmal subarachnoid hemorrhage (aSAH). In this study, we aimed to determine whether lower ICPV also correlated with worse cerebral energy metabolism after aSAH.

METHODS

A total of 75 aSAH patients treated in the neurointensive care unit at Uppsala University Hospital, Sweden between 2008 and 2018 and with both intracranial pressure and cerebral microdialysis (MD) monitoring during the first 10 days after ictus were included in this retrospective study. ICPV was calculated with a bandpass filter limited to intracranial pressure slow waves with a wavelength of 55 to 15 seconds. Cerebral energy metabolites were measured hourly with MD. The monitoring period was divided into 3 phases; early (days 1 to 3), early vasospasm (days 4 to 6.5), and late vasospasm (days 6.5 to 10).

RESULTS

Lower ICPV was associated with lower MD-glucose in the late vasospasm phase, lower MD-pyruvate in the early vasospasm phases, and higher MD-lactate-pyruvate ratio (LPR) in the early and late vasospasm phases. Lower ICPV was associated with poor cerebral substrate supply (LPR >25 and pyruvate <120 µM) rather than mitochondrial failure (LPR >25 and pyruvate >120 µM). There was no association between ICPV and delayed ischemic neurological deficit, but lower ICPV in both vasospasm phases correlated with unfavorable outcomes.

CONCLUSION

Lower ICPV was associated with an increased risk for disturbed cerebral energy metabolism and worse clinical outcomes in aSAH patients, possibly explained by a vasospasm-related decrease in cerebral blood volume dynamics and cerebral ischemia.

Higher intracranial pressure variability is associated with lower cerebrovascular resistance in aneurysmal subarachnoid hemorrhage

Higher intracranial pressure variability (ICPV) has been associated with a more favorable cerebral energy metabolism, lower rate of delayed ischemic neurologic deficits, and more favorable outcome in aneurysmal subarachnoid hemorrhage (aSAH). We have hypothesized that higher ICPV partly reflects more compliant and active cerebral vessels. In this study, the aim was to further test this by investigating if higher ICPV was associated with lower cerebrovascular resistance (CVR) and higher cerebral blood flow (CBF) after aSAH. In this observational study, 147 aSAH patients were included, all of whom had been treated in the Neurointensive Care (NIC) Unit, Uppsala, Sweden, 2012-2020. They were required to have had ICP monitoring and at least one xenon-enhanced computed tomography (Xe-CT) scan to study cortical CBF within the first 2 weeks post-ictus. CVR was defined as the cerebral perfusion pressure in association with the Xe-CT scan divided by the concurrent CBF. ICPV was defined over three intervals: subminute (ICPV-1m), 30-min (ICPV-30m), and 4 h (ICPV-4h). The first 14 days were divided into early (days 1-3) and vasospasm phase (days 4-14). In the vasospasm phase, but not in the early phase, higher ICPV-4h (β =  - 0.19, p < 0.05) was independently associated with a lower CVR in a multiple linear regression analysis and with a higher global cortical CBF (r = 0.19, p < 0.05) in a univariate analysis. ICPV-1m and ICPV-30m were not associated with CVR or CBF in any phase. This study corroborates the hypothesis that higher ICPV, at least in the 4-h interval, is favorable and may reflect more compliant and possibly more active cerebral vessels.

Permutation Entropy Analysis to Intracranial Hypertension from a Porcine Model

Intracranial pressure (ICP) monitoring is commonly used in the follow-up of patients in intensive care units, but only a small part of the information available in the ICP time series is exploited. One of the most important features to guide patient follow-up and treatment is intracranial compliance. We propose using permutation entropy (PE) as a method to extract non-obvious information from the ICP curve. We analyzed the results of a pig experiment with sliding windows of 3600 samples and 1000 displacement samples, and estimated their respective PEs, their associated probability distributions, and the number of missing patterns (NMP). We observed that the behavior of PE is inverse to that of ICP, in addition to the fact that NMP appears as a surrogate for intracranial compliance. In lesion-free periods, PE is usually greater than 0.3, and normalized NMP is less than 90% and p(s1)>p(s720). Any deviation from these values could be a possible warning of altered neurophysiology. In the terminal phases of the lesion, the normalized NMP is higher than 95%, and PE is not sensitive to changes in ICP and p(s720)>p(s1). The results show that it could be used for real-time patient monitoring or as input for a machine learning tool.

Low intracranial pressure variability is associated with delayed cerebral ischemia and unfavorable outcome in aneurysmal subarachnoid hemorrhage

Purpose

High intracranial pressure variability (ICPV) is associated with favorable outcome in traumatic brain injury, by mechanisms likely involving better cerebral blood flow regulation. However, less is known about ICPV in aneurysmal subarachnoid hemorrhage (aSAH). In this study, we investigated the explanatory variables for ICPV in aSAH and its association with delayed cerebral ischemia (DCI) and clinical outcome.

Methods

In this retrospective study, 242 aSAH patients, treated at the neurointensive care, Uppsala, Sweden, 2008–2018, with ICP monitoring the first ten days post-ictus were included. ICPV was evaluated on three time scales: (1) ICPV-1 m—ICP slow wave amplitude of wavelengths between 55 and 15 s, (2) ICPV-30 m—the deviation from the mean ICP averaged over 30 min, and (3) ICPV-4 h—the deviation from the mean ICP averaged over 4 h. The ICPV measures were analyzed in the early phase (day 1–3), in the early vasospasm phase (day 4–6.5), and the late vasospasm phase (day 6.5–10).

Results

High ICPV was associated with younger age, reduced intracranial pressure/volume reserve (high RAP), and high blood pressure variability in multiple linear regression analyses for all ICPV measures. DCI was associated with reduced ICPV in both vasospasm phases. High ICPV-1 m in the post-ictal early phase and the early vasospasm phase predicted favorable outcome in multiple logistic regressions, whereas ICPV-30 m and ICPV-4 h in the late vasospasm phase had a similar association.

Conclusions

Higher ICPV may reflect more optimal cerebral vessel activity, as reduced values are associated with an increased risk of DCI and unfavorable outcome after aSAH.

Association between Physiological Signal Complexity and Outcomes in Moderate and Severe Traumatic Brain Injury: A CENTER-TBI Exploratory Analysis of Multi-Scale Entropy

In traumatic brain injury (TBI), preliminary retrospective work on signal entropy suggests an association with global outcome. The goal of this study was to provide multi-center validation of the association between multi-scale entropy (MSE) of cardiovascular and cerebral physiological signals, with six-month outcome. Using the Collaborative European NeuroTrauma Effectiveness Research in TBI (CENTER-TBI) high-resolution intensive care unit (ICU) cohort, we selected patients with a minimum of 72 h of physiological recordings and a documented six-month Glasgow Outcome Scale Extended (GOSE) score. The 10-sec summary data for heart rate (HR), mean arterial pressure (MAP), intracranial pressure (ICP), and pulse amplitude of ICP (AMP) were derived across the first 72 h of data. The MSE complexity index (MSE-Ci) was determined for HR, MAP, ICP, and AMP, with the association between MSE and dichotomized six-month outcomes assessed using Mann-Whitney U testing and logistic regression analysis. A total of 160 patients had a minimum of 72 h of recording and a documented outcome. Decreased HR MSE-Ci (7.3 [interquartile range (IQR) 5.4 to 10.2] vs. 5.1 [IQR 3.1 to 7.0]; p = 0.002), lower ICP MSE-Ci (11.2 [IQR 7.5 to 14.2] vs. 7.3 [IQR 6.1 to 11.0]; p = 0.009), and lower AMP MSE-Ci (10.9 [IQR 8.0 to 13.7] vs. 8.7 [IQR 6.6 to 11.0]; p = 0.022), were associated with death. Similarly, lower HR MSE-Ci (8.0 [IQR 6.2 to 10.9] vs. 6.2 [IQR 3.9 to 8.7]; p = 0.003) and lower ICP MSE-Ci (11.4 [IQR 8.6 to 14.4)] vs. 9.2 [IQR 6.0 to 13.5]), were associated with unfavorable outcome. Logistic regression analysis confirmed that lower HR MSE-Ci and ICP MSE-Ci were associated with death and unfavorable outcome at six months. These findings suggest that a reduction in cardiovascular and cerebrovascular system entropy is associated with worse outcomes. Further work in the field of signal complexity in TBI multi-modal monitoring is required.

Intracranial Pressure Monitoring Signals After Traumatic Brain Injury: A Narrative Overview and Conceptual Data Science Framework

Continuous intracranial pressure (ICP) monitoring is a cornerstone of neurocritical care after severe brain injuries such as traumatic brain injury and acts as a biomarker of secondary brain injury. With the rapid development of artificial intelligent (AI) approaches to data analysis, the acquisition, storage, real-time analysis, and interpretation of physiological signal data can bring insights to the field of neurocritical care bioinformatics. We review the existing literature on the quantification and analysis of the ICP waveform and present an integrated framework to incorporate signal processing tools, advanced statistical methods, and machine learning techniques in order to comprehensively understand the ICP signal and its clinical importance. Our goals were to identify the strengths and pitfalls of existing methods for data cleaning, information extraction, and application. In particular, we describe the use of ICP signal analytics to detect intracranial hypertension and to predict both short-term intracranial hypertension and long-term clinical outcome. We provide a well-organized roadmap for future researchers based on existing literature and a computational approach to clinically-relevant biomedical signal data.

Signal Information Prediction of Mortality Identifies Unique Patient Subsets after Severe Traumatic Brain Injury: A Decision-Tree Analysis Approach

Nonlinear physiological signal features that reveal information content and causal flow have recently been shown to be predictors of mortality after severe traumatic brain injury (TBI). The extent to which these features interact together, and with traditional measures to describe patients in a clinically meaningful way remains unclear. In this study, we incorporated basic demographics (age and initial Glasgow Coma Scale [GCS]) with linear and non-linear signal information based features (approximate entropy [ApEn], and multivariate conditional Granger causality [GC]) to evaluate their relative contributions to mortality using cardio-cerebral monitoring data from 171 severe TBI patients admitted to a single neurocritical care center over a 10 year period. Beyond linear modelling, we employed a decision tree analysis approach to define a predictive hierarchy of features. We found ApEn (p = 0.009) and GC (p = 0.004) based features to be independent predictors of mortality at a time when mean intracranial pressure (ICP) was not. Our combined model with both signal information-based features performed the strongest (area under curve = 0.86 vs. 0.77 for linear features only). Although low "intracranial" complexity (ApEn-ICP) outranked both age and GCS as crucial drivers of mortality (fivefold increase in mortality where ApEn-ICP <1.56, 36.2% vs. 7.8%), decision tree analysis revealed clear subsets of patient populations using all three predictors. Patients with lower ApEn-ICP who were >60 years of age died, whereas those with higher ApEn-ICP and GCS ≥5 all survived. Yet, even with low initial intracranial complexity, as long as patients maintained robust GC and "extracranial" complexity (ApEn of mean arterial pressure), they all survived. Incorporating traditional linear and novel, non-linear signal information features, particularly in a framework such as decision trees, may provide better insight into "health" status. However, caution is required when interpreting these results in a clinical setting prior to external validation.

Non-linear Dynamical Analysis of Intraspinal Pressure Signal Predicts Outcome After Spinal Cord Injury

The injured spinal cord is a complex system influenced by many local and systemic factors that interact over many timescales. To help guide clinical management, we developed a technique that monitors intraspinal pressure from the injury site in patients with acute, severe traumatic spinal cord injuries. Here, we hypothesize that spinal cord injury alters the complex dynamics of the intraspinal pressure signal quantified by computing hourly the detrended fluctuation exponent alpha, multiscale entropy, and maximal Lyapunov exponent lambda. 49 patients with severe traumatic spinal cord injuries were monitored within 72 h of injury for 5 days on average to produce 5,941 h of intraspinal pressure data. We computed the spinal cord perfusion pressure as mean arterial pressure minus intraspinal pressure and the vascular pressure reactivity index as the running correlation coefficient between intraspinal pressure and arterial blood pressure. Mean patient follow-up was 17 months. We show that alpha values are greater than 0.5, which indicates that the intraspinal pressure signal is fractal. As alpha increases, intraspinal pressure decreases and spinal cord perfusion pressure increases with negative correlation between the vascular pressure reactivity index vs. alpha. Thus, secondary insults to the injured cord disrupt intraspinal pressure fractality. Our analysis shows that high intraspinal pressure, low spinal cord perfusion pressure, and impaired pressure reactivity strongly correlate with reduced multi-scale entropy, supporting the notion that secondary insults to the injured cord cause de-complexification of the intraspinal pressure signal, which may render the cord less adaptable to external changes. Healthy physiological systems are characterized by edge of chaos dynamics. We found negative correlations between the percentage of hours with edge of chaos dynamics (−0.01 ≤ lambda ≤ 0.01) vs. high intraspinal pressure and vs. low spinal cord perfusion pressure; these findings suggest that secondary insults render the intraspinal pressure more regular or chaotic. In a multivariate logistic regression model, better neurological status on admission, higher intraspinal pressure multi-scale entropy and more frequent edge of chaos intraspinal pressure dynamics predict long-term functional improvement. We conclude that spinal cord injury is associated with marked changes in non-linear intraspinal pressure metrics that carry prognostic information.

Continuous Monitoring of the Complexity of Intracranial Pressure After Head Injury

Multiscale entropy (MSE) has been increasingly used to investigate the complexity of biological signals. Our previous study demonstrated that the complexity of mean intracranial pressure (ICP), assessed by MSE based on the whole recording periods, is associated with the outcome after traumatic brain injury (TBI). To improve the feasibility of MSE in a clinical setting, this study examined whether the complexity of ICP waveforms based on shorter periods could be a reliable predictor of the outcome in patients with TBI. Results showed that the complexity of ICP slow waves, calculated in 3-h moving windows, correlates with the outcome of patients with TBI. Thus, the complexity of ICP may be a promising index to be incorporated into multimodal monitoring in patients with TBI.

Idiopathic Intracranial Hypertension in Monozygotic Female Twins: Intracranial Pressure Dynamics and Treatment Outcome

BACKGROUND

Familial cases of idiopathic intracranial hypertension (IIH) are exceedingly rare, and its occurrence in monozygotic twins has not been reported previously.

CASE DESCRIPTION

We report monozygotic female twins who developed IIH, one at age 25 years and the other at age 28 years. Continuous intracranial pressure (ICP) monitoring confirmed elevated ICP as measured initially by lumbar puncture. In both cases, successful treatment with resolution of papilledema and symptoms relief was achieved after ventriculoperitoneal shunting.

CONCLUSIONS

This report documents the first case of IIH in monozygotic twins and the associated changes in ICP dynamics. Interestingly, almost equivalent alterations in ICP dynamics were found in the 2 patients.

Effect of salt intake on beat-to-beat blood pressure nonlinear dynamics and entropy in salt-sensitive versus salt-protected rats

Blood pressure exhibits substantial short- and long-term variability (BPV). We assessed the hypothesis that the complexity of beat-to-beat BPV will be differentially altered in salt-sensitive hypertensive Dahl rats (SS) versus rats protected from salt-induced hypertension (SSBN13) maintained on high-salt versus low-salt diet. Beat-to-beat systolic and diastolic BP series from nine SS and six SSBN13 rats (http://www.physionet.org) were analyzed following 9 weeks on low salt and repeated after 2 weeks on high salt. BP complexity was quantified by detrended fluctuation analysis (DFA), short- and long-range scaling exponents (αS and αL), sample entropy (SampEn), and traditional standard deviation (SD) and coefficient of variation (CV(%)). Mean systolic and diastolic BP increased on high-salt diet (P < 0.01) particularly for SS rats. SD and CV(%) were similar across groups irrespective of diet. Salt-sensitive and -protected rats exhibited similar complexity indices on low-salt diet. On high salt, (1) SS rats showed increased scaling exponents or smoother, systolic (P = 0.007 [αL]) and diastolic (P = 0.008 [αL]) BP series; (2) salt-protected rats showed lower SampEn (less complex) systolic and diastolic BP (P = 0.046); and (3) compared to protected SSBN13 rats, SS showed higher αL for systolic (P = 0.01) and diastolic (P = 0.005) BP Hypertensive SS rats are more susceptible to high salt with a greater rise in mean BP and reduced complexity. Comparable mean pressures in sensitive and protective rats when on low-salt diet coupled with similar BPV dynamics suggest a protective role of low-salt intake in hypertensive rats. This effect likely reflects better coupling of biologic oscillators.

Intracranial pressure for the characterization of different types of hydrocephalus: A Permutation Entropy study

Hydrocephalus is a condition characterized by altered cerebrospinal fluid (CSF) dynamics and chronic rises in intracranial pressure (ICP). However, the reason why hydrocephalic physiologies fail to inhibit dangerously high ICP levels is not known. Infusion studies are used to raise ICP and evaluate CSF circulation disorders. In this pilot study, ICP signals recorded during infusion tests from 33 patients with normal pressure hydrocephalus and 36 patients having developed a secondary form of normal pressure hydrocephalus were characterized using Permutation Entropy (PE), a symbolic non-linear method to quantify complexity. Each ICP signal was divided into four epochs--baseline (before infusion begins), infusion, plateau, and recovery (after infusion has stopped)--and the mean PE was calculated for each epoch. Statistically significant differences were found between PE for most epochs (p<;0.00833, Bonferroni-corrected Wilcoxon tests), with a significant decrease in the plateau phase. However, differences between PE for normal pressure and secondary hydrocephalus were not significant. Results suggest that the increase in ICP during infusion studies is associated with a significant decrease in PE. PE analysis of ICP signals could be useful for increasing our understanding of CSF dynamics in normal pressure hydrocephalus.