Automatic identification of intracranial pressure waveform during external ventricular drainage clamping: segmentation via wavelet analysis.

Objective. The objective of this study is to develop and validate a method for automatically identifying segments of intracranial pressure (ICP) waveform data from external ventricular drainage (EVD) recordings during intermittent drainage and closure.Methods. The proposed method uses time-frequency analysis through wavelets to distinguish periods of ICP waveform in EVD data. By comparing the frequency compositions of the ICP signals (when the EVD system is clamped) and the artifacts (when the system is open), the algorithm can detect short, uninterrupted segments of ICP waveform from the longer periods of non-measurement data. The method involves applying a wavelet transform, calculating the absolute power in a specific range, using Otsu thresholding to automatically identify a threshold, and performing a morphological operation to remove small segments. Two investigators manually graded the same randomly selected one-hour segments of the resulting processed data. Performance metrics were calculated as a percentage.Results. The study analyzed data from 229 patients who had EVD placed following subarachnoid hemorrhage between June 2006 and December 2012. Of these, 155 (67.7%) were female and 62 (27%) developed delayed cerebral ischemia. A total of 45 150 h of data were segmented. 2044 one-hour segments were randomly selected and evaluated by two investigators (MM and DN). Of those, the evaluators agreed on the classification of 1556 one-hour segments. The algorithm was able to correctly identify 86% (1338 h) of ICP waveform data. 8.2% (128 h) of the time the algorithm either partially or fully failed to segment the ICP waveform. 5.4% (84 h) of data, artifacts were mistakenly identified as ICP waveforms (false positives).Conclusion. The proposed algorithm automates the identification of valid ICP waveform segments of waveform in EVD data and thus enables the inclusion in real-time data analysis for decision support. It also standardizes and makes research data management more efficient.

A Deep Learning Framework for Deriving Noninvasive Intracranial Pressure Waveforms from Transcranial Doppler.

Increased intracranial pressure (ICP) causes disability and mortality in the neurointensive care population. Current methods for monitoring ICP are invasive. We designed a deep learning framework using a domain adversarial neural network to estimate noninvasive ICP, from blood pressure, electrocardiogram, and cerebral blood flow velocity. Our model had a mean of median absolute error of 3.88 ± 3.26 mmHg for the domain adversarial neural network, and 3.94 ± 1.71 mmHg for the domain adversarial transformers. Compared with nonlinear approaches, such as support vector regression, this was 26.7% and 25.7% lower. Our proposed framework provides more accurate noninvasive ICP estimates than currently available. ANN NEUROL 2023;94:196-202.

Optimal Cerebral Perfusion Pressure and Brain Tissue Oxygen in Aneurysmal Subarachnoid Hemorrhage.

BACKGROUND: Targeting a cerebral perfusion pressure optimal for cerebral autoregulation (CPPopt) has been gaining more attention to prevent secondary damage after acute neurological injury. Brain tissue oxygenation (PbtO2) can identify insufficient cerebral blood flow and secondary brain injury. Defining the relationship between CPPopt and PbtO2 after aneurysmal subarachnoid hemorrhage may result in (1) mechanistic insights into whether and how CPPopt-based strategies might be beneficial and (2) establishing support for the use of PbtO2 as an adjunctive monitor for adequate or optimal local perfusion. METHODS: We performed a retrospective analysis of a prospectively collected 2-center dataset of patients with aneurysmal subarachnoid hemorrhage with or without later diagnosis of delayed cerebral ischemia (DCI). CPPopt was calculated as the cerebral perfusion pressure (CPP) value corresponding to the lowest pressure reactivity index (moving correlation coefficient of mean arterial and intracranial pressure). The relationship of (hourly) deltaCPP (CPP-CPPopt) and PbtO2 was investigated using natural spline regression analysis. Data after DCI diagnosis were excluded. Brain tissue hypoxia was defined as PbtO2 <20 mmHg. RESULTS: One hundred thirty-one patients were included with a median of 44.0 (interquartile range, 20.8-78.3) hourly CPPopt/PbtO2 datapoints. The regression plot revealed a nonlinear relationship between PbtO2 and deltaCPP (P<0.001) with PbtO2 decrease with deltaCPP <0 mmHg and stable PbtO2 with deltaCPP ≥0mmHg, although there was substantial individual variation. Brain tissue hypoxia (34.6% of all measurements) was more frequent with deltaCPP <0 mmHg. These dynamics were similar in patients with or without DCI. CONCLUSIONS: We found a nonlinear relationship between PbtO2 and deviation of patients' CPP from CPPopt in aneurysmal subarachnoid hemorrhage patients in the pre-DCI period. CPP values below calculated CPPopt were associated with lower PbtO2. Nevertheless, the nature of PbtO2 measurements is complex, and the variability is high. Combined multimodality monitoring with CPP/CPPopt and PbtO2 should be recommended to redefine individual pressure targets (CPP/CPPopt) and retain the option to detect local perfusion deficits during DCI (PbtO2), which cannot be fulfilled by both measurements interchangeably.

Heart rate variability and adrenal size provide clues to sudden cardiac death in hospitalized COVID-19 patients.

PURPOSE: To examine the association between a measure of heart rate variability and sudden cardiac death (SCD) in COVID-19 patients. METHODS: Patients with SARS-COV-2 infection admitted to Columbia University Irving Medical Center who died between 4/25/2020 and 7/14/2020 and had an autopsy were examined for root mean square of successive differences (RMSSD), organ weights, and evidence of SCD. RESULTS: Thirty COVID-19 patients were included and 12 had SCD. The RMSSD over 7 days without vs with SCD was median 0.0129 (IQR 0.0074-0.026) versus 0.0098 (IQR 0.0056-0.0197), p < 0.0001. The total adjusted adrenal weight of the non-SCD group was 0.40 g/kg (IQR 0.35-0.55) versus 0.25 g/kg (IQR 0.21-0.31) in the SCD group, p = 0.0007. CONCLUSIONS: Hospitalized patients with COVID-19 who experienced SCD had lower parasympathetic activity (RMSSD) and smaller sized adrenal glands. Further research is required to replicate these findings.

Predicting Shunt Dependency from the Effect of Cerebrospinal Fluid Drainage on Ventricular Size.

BACKGROUND: Prolonged external ventricular drainage (EVD) in patients with subarachnoid hemorrhage (SAH) leads to morbidity, whereas early removal can have untoward effects related to recurrent hydrocephalus. A metric to help determine the optimal time for EVD removal or ventriculoperitoneal shunt (VPS) placement would be beneficial in preventing the prolonged, unnecessary use of EVD. This study aimed to identify whether dynamics of cerebrospinal fluid (CSF) biometrics can temporally predict VPS dependency after SAH. METHODS: This was a retrospective analysis of a prospective, single-center, observational study of patients with aneurysmal SAH who required EVD placement for hydrocephalus. Patients were divided into VPS-dependent (VPS+) and non-VPS dependent groups. We measured the bicaudate index (BCI) on all available computed tomography scans and calculated the change over time (ΔBCI). We analyzed the relationship of ΔBCI with CSF output by using Pearson's correlation. A k-nearest neighbor model of the relationship between ΔBCI and CSF output was computed to classify VPS. RESULTS: Fifty-eight patients met inclusion criteria. CSF output was significantly higher in the VPS+ group in the 7 days post EVD placement. There was a negative correlation between delta BCI and CSF output in the VPS+ group (negative delta BCI means ventricles become smaller) and a positive correlation in the VPS- group starting from days four to six after EVD placement (p < 0.05). A weighted k-nearest neighbor model for classification had a sensitivity of 0.75, a specificity of 0.70, and an area under the receiver operating characteristic curve of 0.80. CONCLUSIONS: The correlation of ΔBCI and CSF output is a reliable intraindividual biometric for VPS dependency after SAH as early as days four to six after EVD placement. Our machine learning model leverages this relationship between ΔBCI and cumulative CSF output to predict VPS dependency. Early knowledge of VPS dependency could be studied to reduce EVD duration in many centers (intensive care unit length of stay).

Dynamic Intracranial Pressure Waveform Morphology Predicts Ventriculitis.

BACKGROUND: Intracranial pressure waveform morphology reflects compliance, which can be decreased by ventriculitis. We investigated whether morphologic analysis of intracranial pressure dynamics predicts the onset of ventriculitis. METHODS: Ventriculitis was defined as culture or Gram stain positive cerebrospinal fluid, warranting treatment. We developed a pipeline to automatically isolate segments of intracranial pressure waveforms from extraventricular catheters, extract dominant pulses, and obtain morphologically similar groupings. We used a previously validated clinician-supervised active learning paradigm to identify metaclusters of triphasic, single-peak, or artifactual peaks. Metacluster distributions were concatenated with temperature and routine blood laboratory values to create feature vectors. A L2-regularized logistic regression classifier was trained to distinguish patients with ventriculitis from matched controls, and the discriminative performance using area under receiver operating characteristic curve with bootstrapping cross-validation was reported. RESULTS: Fifty-eight patients were included for analysis. Twenty-seven patients with ventriculitis from two centers were identified. Thirty-one patients with catheters but without ventriculitis were selected as matched controls based on age, sex, and primary diagnosis. There were 1590 h of segmented data, including 396,130 dominant pulses in patients with ventriculitis and 557,435 pulses in patients without ventriculitis. There were significant differences in metacluster distribution comparing before culture-positivity versus during culture-positivity (p < 0.001) and after culture-positivity (p < 0.001). The classifier demonstrated good discrimination with median area under receiver operating characteristic 0.70 (interquartile range 0.55-0.80). There were 1.5 true alerts (ventriculitis detected) for every false alert. CONCLUSIONS: Intracranial pressure waveform morphology analysis can classify ventriculitis without cerebrospinal fluid sampling.

Identification of Endotypes of Hospitalized COVID-19 Patients.

Background: Characterization of coronavirus disease 2019 (COVID-19) endotypes may help explain variable clinical presentations and response to treatments. While risk factors for COVID-19 have been described, COVID-19 endotypes have not been elucidated. Objectives: We sought to identify and describe COVID-19 endotypes of hospitalized patients. Methods: Consensus clustering (using the ensemble method) of patient age and laboratory values during admission identified endotypes. We analyzed data from 528 patients with COVID-19 who were admitted to telemetry capable beds at Columbia University Irving Medical Center and discharged between March 12 to July 15, 2020. Results: Four unique endotypes were identified and described by laboratory values, demographics, outcomes, and treatments. Endotypes 1 and 2 were comprised of low numbers of intubated patients (1 and 6%) and exhibited low mortality (1 and 6%), whereas endotypes 3 and 4 included high numbers of intubated patients (72 and 85%) with elevated mortality (21 and 43%). Endotypes 2 and 4 had the most comorbidities. Endotype 1 patients had low levels of inflammatory markers (ferritin, IL-6, CRP, LDH), low infectious markers (WBC, procalcitonin), and low degree of coagulopathy (PTT, PT), while endotype 4 had higher levels of those markers. Conclusions: Four unique endotypes of hospitalized patients with COVID-19 were identified, which segregated patients based on inflammatory markers, infectious markers, evidence of end-organ dysfunction, comorbidities, and outcomes. High comorbidities did not associate with poor outcome endotypes. Further work is needed to validate these endotypes in other cohorts and to study endotype differences to treatment responses.

Use of Clustering to Investigate Changes in Intracranial Pressure Waveform Morphology in Patients with Ventriculitis.

OBJECTIVE: This study aimed to examine whether changes in intracranial pressure (ICP) waveform morphologies can be used as a biomarker for early detection of ventriculitis. METHODS: Consecutive patients (N = 1653) were prospectively enrolled in a hemorrhage outcomes study from 2006 to 2018. Of these, 435 patients (26%) required external ventricular drains (EVDs) and 76 (17.5% of those with EVDs) had ventriculitis treated with antibiotics. Nineteen patients (25% of those with ventriculitis) showed culture-positive cerebrospinal fluid (CSF) and were included in the present analysis. CSF was routinely cultured three times per week and additionally if infection was suspected. EVDs were left open for drainage, with ICP assessed hourly by clamping. Using wavelet analysis, we extracted uninterrupted segments of ICP waveforms. We extracted dominant pulses from continuous high-resolution data, using morphological clustering analysis of intracranial pressure (MOCAIP). Then we applied k-means clustering, using the dynamic time warping distance to obtain morphologically similar groupings. Finally, metaclusters and further-split clusters (when equipoise existed) were categorized for broad comparison by clinician consensus. RESULTS: We extracted 275,911 dominant pulses from 459.9 h of EVD data. Of these, 112,898 pulses (40.9%) occurred before culture positivity, 41,300 pulses (15.0%) occurred during culture positivity, and 121,713 pulses (44.1%) occurred after it. K-means identified 20 clusters, which were further grouped into metaclusters: tri-/biphasic, single-peak, and artifactual waveforms. Prior to ventriculitis, 61.8% of dominant pulses were tri-/biphasic; this percentage reduced to 22.6% during ventriculitis and 28.4% after it (p < 0.0001). One day before the first positive cultures were collected, the distribution of metaclusters changed to include more single-peak and artifactual ICP waveforms (p < 0.0001). CONCLUSION: The distribution of ICP waveform morphology changes significantly prior to clinical diagnosis of ventriculitis and may be a potential biomarker.

Dynamic Detection of Delayed Cerebral Ischemia: A Study in 3 Centers.

BACKGROUND AND PURPOSE: Delayed cerebral ischemia (DCI) after aneurysmal subarachnoid hemorrhage negatively impacts long-term recovery but is often detected too late to prevent damage. We aim to develop hourly risk scores using routinely collected clinical data to detect DCI. METHODS: A DCI classification model was trained using vital sign measurements (heart rate, blood pressure, respiratory rate, and oxygen saturation) and demographics routinely collected for clinical care. Twenty-two time-varying physiological measures were computed including mean, SD, and cross-correlation of heart rate time series with each of the other vitals. Classification was achieved using an ensemble approach with L2-regularized logistic regression, random forest, and support vector machines models. Classifier performance was determined by area under the receiver operating characteristic curves and confusion matrices. Hourly DCI risk scores were generated as the posterior probability at time t using the Ensemble classifier on cohorts recruited at 2 external institutions (n=38 and 40). RESULTS: Three hundred ten patients were included in the training model (median, 54 years old [interquartile range, 45-65]; 80.2% women, 28.4% Hunt and Hess scale 4-5, 38.7% Modified Fisher Scale 3-4); 101 (33%) developed DCI with a median onset day 6 (interquartile range, 5-8). Classification accuracy before DCI onset was 0.83 (interquartile range, 0.76-0.83) area under the receiver operating characteristic curve. Risk scores applied to external institution datasets correctly predicted 64% and 91% of DCI events as early as 12 hours before clinical detection, with 2.7 and 1.6 true alerts for every false alert. CONCLUSIONS: An hourly risk score for DCI derived from routine vital signs may have the potential to alert clinicians to DCI, which could reduce neurological injury.

Hyperemia in subarachnoid hemorrhage patients is associated with an increased risk of seizures.

The association between impaired brain perfusion, cerebrovascular reactivity status and the risk of ictal events in patients with subarachnoid hemorrhage is unknown. We identified 13 subarachnoid hemorrhage (SAH) patients with seizures and 22 with ictal-interictal continuum (IIC), and compared multimodality physiological recordings to 38 similarly poor-grade SAH patients without ictal activity. We analyzed 10,179 cumulative minutes of seizure and 12,762 cumulative minutes of IIC. Cerebrovascular reactivity (PRx) was not different between subjects with seizures, IIC, or controls. Cerebral perfusion pressure (CPP) was higher in patients with seizures [99 ± 6.5, p = .005] and IIC [97 ± 8.5, p = .007] when compared to controls [89 ± 12.3]. DeltaCPP, defined as actual CPP minus optimal CPP (CPPopt), was also higher in the seizure group [8.3 ± 7.9, p = .0003] and IIC [8.1 ± 10.3, p = .0006] when compared to controls [-0.1 ± 5]. Time spent with supra-optimal CPP was higher in the seizure group [342 ± 213 min/day, p = .002] when compared to controls [154 ± 120 min/day]. In a temporal examination, a supra-optimal CPP preceded increased seizures and IIC in SAH patients, an hour before and continued to increase during the events [p < .0001].

Heart Rate Variability as a Biomarker of Neurocardiogenic Injury After Subarachnoid Hemorrhage.

BACKGROUND: The objective of this study was to examine whether heart rate variability (HRV) measures can be used to detect neurocardiogenic injury (NCI). METHODS: Three hundred and twenty-six consecutive admissions with aneurysmal subarachnoid hemorrhage (SAH) met criteria for the study. Of 326 subjects, 56 (17.2%) developed NCI which we defined by wall motion abnormality with ventricular dysfunction on transthoracic echocardiogram or cardiac troponin-I > 0.3 ng/mL without electrocardiogram evidence of coronary artery insufficiency. HRV measures (in time and frequency domains, as well as nonlinear technique of detrended fluctuation analysis) were calculated over the first 48 h. We applied longitudinal multilevel linear regression to characterize the relationship of HRV measures with NCI and examine between-group differences at baseline and over time. RESULTS: There was decreased vagal activity in NCI subjects with a between-group difference in low/high frequency ratio (ß 3.42, SE 0.92, p = 0.0002), with sympathovagal balance in favor of sympathetic nervous activity. All time-domain measures were decreased in SAH subjects with NCI. An ensemble machine learning approach translated these measures into a classification tool that demonstrated good discrimination using the area under the receiver operating characteristic curve (AUROC 0.82), the area under precision recall curve (AUPRC 0.75), and a correct classification rate of 0.81. CONCLUSIONS: HRV measures are significantly associated with our label of NCI and a machine learning approach using features derived from HRV measures can classify SAH patients that develop NCI.

Predicting delayed cerebral ischemia after subarachnoid hemorrhage using physiological time series data.

To develop and validate a prediction model for delayed cerebral ischemia (DCI) after subarachnoid hemorrhage (SAH) using a temporal unsupervised feature engineering approach, demonstrating improved precision over standard features. 488 consecutive SAH admissions from 2006 to 2014 to a tertiary care hospital were included. Models were trained on 80%, while 20% were set aside for validation testing. Baseline information and standard grading scales were evaluated: age, sex, Hunt Hess grade, modified Fisher Scale (mFS), and Glasgow Coma Scale (GCS). An unsupervised approach applying random kernels was used to extract features from physiological time series (systolic and diastolic blood pressure, heart rate, respiratory rate, and oxygen saturation). Classifiers (Partial Least Squares, linear and kernel Support Vector Machines) were trained on feature subsets of the derivation dataset. Models were applied to the validation dataset. The performances of the best classifiers on the validation dataset are reported by feature subset. Standard grading scale (mFS): AUC 0.58. Combined demographics and grading scales: AUC 0.60. Random kernel derived physiologic features: AUC 0.74. Combined baseline and physiologic features with redundant feature reduction: AUC 0.77. Current DCI prediction tools rely on admission imaging and are advantageously simple to employ. However, using an agnostic and computationally inexpensive learning approach for high-frequency physiologic time series data, we demonstrated that our models achieve higher classification accuracy.

An active learning framework for enhancing identification of non-artifactual intracranial pressure waveforms.

OBJECTIVE: Intracranial pressure (ICP) is an important and established clinical measurement that is used in the management of severe acute brain injury. ICP waveforms are usually triphasic and are susceptible to artifact because of transient catheter malfunction or routine patient care. Existing methods for artifact detection include threshold-based, stability-based, or template matching, and result in higher false positives (when there is variability in the ICP waveforms) or higher false negatives (when the ICP waveforms lack complete triphasic components but are valid). APPROACH: We hypothesized that artifact labeling of ICP waveforms can be optimized by an active learning approach which includes interactive querying of domain experts to identify a manageable number of informative training examples. MAIN RESULTS: The resulting active learning based framework identified non-artifactual ICP pulses with a superior AUC of 0.96 + 0.012, compared to existing methods: template matching (AUC: 0.71 + 0.04), ICP stability (AUC: 0.51 + 0.036) and threshold-based (AUC: 0.5 + 0.02). SIGNIFICANCE: The proposed active learning framework will support real-time ICP-derived analytics by improving precision of artifact-labelling.

Incorporating High-Frequency Physiologic Data Using Computational Dictionary Learning Improves Prediction of Delayed Cerebral Ischemia Compared to Existing Methods.

PURPOSE: Accurate prediction of delayed cerebral ischemia (DCI) after subarachnoid hemorrhage (SAH) can be critical for planning interventions to prevent poor neurological outcome. This paper presents a model using convolution dictionary learning to extract features from physiological data available from bedside monitors. We develop and validate a prediction model for DCI after SAH, demonstrating improved precision over standard methods alone. METHODS: 488 consecutive SAH admissions from 2006 to 2014 to a tertiary care hospital were included. Models were trained on 80%, while 20% were set aside for validation testing. Modified Fisher Scale was considered the standard grading scale in clinical use; baseline features also analyzed included age, sex, Hunt-Hess, and Glasgow Coma Scales. An unsupervised approach using convolution dictionary learning was used to extract features from physiological time series (systolic blood pressure and diastolic blood pressure, heart rate, respiratory rate, and oxygen saturation). Classifiers (partial least squares and linear and kernel support vector machines) were trained on feature subsets of the derivation dataset. Models were applied to the validation dataset. RESULTS: The performances of the best classifiers on the validation dataset are reported by feature subset. Standard grading scale (mFS): AUC 0.54. Combined demographics and grading scales (baseline features): AUC 0.63. Kernel derived physiologic features: AUC 0.66. Combined baseline and physiologic features with redundant feature reduction: AUC 0.71 on derivation dataset and 0.78 on validation dataset. CONCLUSION: Current DCI prediction tools rely on admission imaging and are advantageously simple to employ. However, using an agnostic and computationally inexpensive learning approach for high-frequency physiologic time series data, we demonstrated that we could incorporate individual physiologic data to achieve higher classification accuracy.

Deriving the PRx and CPPopt from 0.2-Hz Data: Establishing Generalizability to Bedmaster Users.

OBJECTIVE: The objective was to explore the validity of industry-parameterized vital signs in the generation of pressure reactivity index (PRx) and optimal cerebral perfusion pressure (CPPopt) values. MATERIALS AND METHODS: Ten patients with intracranial pressure (ICP) monitors from 2008 to 2013 in a tertiary care hospital were included. Arterial blood pressure (ABP) and ICP were sampled at 240 Hz (of waveform data) and 0.2 Hz (of parameterized data produced by heuristic industry proprietary algorithms). 240-Hz ABP were filtered for pulse pressure and diastolic ABP within the limits of 20-150 mmHg. The PRx was calculated as Pearson's correlation coefficient using 10-s averages of ICP and ABP over a 5-min moving window with 80% overlap. For ease of comparison, we used the naming convention of BMx for PRx values derived from 0.2-Hz data. A 5-min median cerebral perfusion pressure (CPP) trend was calculated, PRx or BMx values divided and averaged into CPP bins spanning 5 mmHg. The minimum Y value (PRx or BMx) of the parabolic function fit to the resulting XY plot of 4 h of data was obtained, and updated every 1 min. Pearson's R correlations were calculated for each patient. Linear mixed-effects models were used with a random intercept to assess the overall correlation between the PRx (outcome) and the BMx (fixed effect) or the CPPopt-PRx (outcome) and the CPPopt-BMx (fixed effect). RESULTS: The overall correlation between the PRx and BMx was 0.78 based on the linear mixed effects models (p < 0.0001), and the overall correlation for the CPPopt-PRx and CPPopt-BMx based on the linear mixed effects models was 0.76 (p < 0.0001). One patient had low correlation of CPPopts derived from the PRx vs the BMx; this patient had the least number of hours of CPPopt data to compare. CONCLUSIONS: The BMx shows promise in CPPopt derivation against the validated PRx measure. If further developed, it could expand the capability of centers to derive CPPopt goals for use in clinical trials.