Peng Xu

Morphological Clustering and Analysis of Continuous Intracranial Pressure

The continuous measurement of intracranial pressure (ICP) is an important and established clinical tool that is used in the management of many neurosurgical disorders such as traumatic brain injury. Only mean ICP information is used currently in the prevailing clinical practice, ignoring the useful information in ICP pulse waveform that can be continuously acquired and is potentially useful for forecasting intracranial and cerebrovascular pathophysiological changes. The present study introduces and validates an algorithm of performing automated analysis of continuous ICP pulse waveform. This algorithm is capable of enhancing ICP signal quality, recognizing non artifactual ICP pulses, and optimally designating the three well-established subcomponents in an ICP pulse. Validation of the proposed algorithm is done by comparing non artifactual pulse recognition and peak designation results from a human observer with those from automated analysis based on a large signal database built from 700 h of recordings from 66 neurosurgical patients. An accuracy of 97.84% is achieved in recognizing non artifactual ICP pulses. An accuracy of 90.17%, 87.56%, and 86.53% was obtained for designating each of the three established ICP subpeaks. These results show that the proposed algorithm can be reliably applied to process continuous ICP recordings from real clinical environment to extract useful morphological features of ICP pulses.

Forecasting ICP Elevation Based on Prescient Changes of Intracranial Pressure Waveform Morphology

Interventions of intracranial pressure (ICP) elevation in neurocritical care is currently delivered only after healthcare professionals notice sustained and significant mean ICP elevation. This paper uses the morphological clustering and analysis of ICP (MOCAIP) algorithm to derive 24 metrics characterizing morphology of ICP pulses and test the hypothesis that preintracranial hypertension (Pre-IH) segments of ICP can be differentiated, using these morphological metrics, from control segments that were not associated with any ICP elevation or at least 1 h prior to ICP elevation. Furthermore, we investigate whether a global optimization algorithm could effectively find the optimal subset of these morphological metrics to achieve better classification performance as compared to using full set of MOCAIP metrics. The results showed that Pre-IH segments, using the optimal subset of metrics found by the differential evolution algorithm, can be differentiated from control segments at a specificity of 99% and sensitivity of 37% for these Pre-IH segments 5 min prior to the ICP elevation. While the sensitivity decreased to 21% for Pre-IH segments, 20 min prior to ICP elevation, the high specificity of 99% was retained. The performance using the full set of MOCAIP metrics was shown inferior to results achieved using the optimal subset of metrics. This paper demonstrated that advanced ICP pulse analysis combined with machine learning could potentially leads to the forecasting of ICP elevation so that a proactive ICP management could be realized based on these accurate forecasts.

An algorithm for extracting intracranial pressure latency relative to electrocardiogram R wave

Intracranial pressure (ICP) latency is defined as the time interval between the peak of the QRS complex of the electrocardiogram (ECG) and the corresponding onset of intracranial pressure (ICP) pulse. Due to its inherent relationship with arterial pulse wave velocity, ICP latency may allow continuous monitoring of pathophysiological changes in the cerebrovasculature. The objective of the present work was to develop and validate a computerized algorithm for extracting ICP latency in a beat-by-beat fashion. The proposed ICP latency extraction algorithm exploits the mature technique of ECG QRS detection and includes a new adaptive peak detection methodology. The results were validated by comparing the performance of two human observers versus the algorithm in terms of locating the onset points of ICP pulses for 59 recordings extracted from 25 adult patients. The average ICP latency was 72.6+/-19.5 ms (range 40.0-159.8). The ICP pulse detection algorithm demonstrated a baseline sensitivity of 0.97 and a positive predictivity of 0.88. No difference was found in the mean location errors from comparing the results obtained by the two observers and those from comparing the results from the algorithm to those from the two observers. Further investigation is needed to demonstrate the role of ICP latency in characterizing dynamic cerebral vascular pathophysiological changes in clinical states such as subarachnoid hemorrhage and traumatic brain injury.

Regression analysis for peak designation in pulsatile pressure signals

Following recent studies, the automatic analysis of intracranial pressure (ICP) pulses appears to be a promising tool for forecasting critical intracranial and cerebrovascular pathophysiological variations during the management of many disorders. A pulse analysis framework has been recently developed to automatically extract morphological features of ICP pulses. The algorithm is able to enhance the quality of ICP signals, to segment ICP pulses, and to designate the locations of the three ICP sub-peaks in a pulse. This paper extends this algorithm by utilizing machine learning techniques to replace Gaussian priors used in the peak designation process with more versatile regression models. The experimental evaluations are conducted on a database of ICP signals built from 700 h of recordings from 64 neurosurgical patients. A comparative analysis of different state-of-the-art regression analysis methods is conducted and the best approach is then compared to the original pulse analysis algorithm. The results demonstrate a significant improvement in terms of accuracy in favor of our regression-based recognition framework. It reaches an average peak designation accuracy of 99% using a kernel spectral regression against 93% for the original algorithm.

A data mining framework for time series estimation

Time series estimation techniques are usually employed in biomedical research to derive variables less accessible from a set of related and more accessible variables. These techniques are traditionally built from systems modeling approaches including simulation, blind decovolution, and state estimation. In this work, we define target time series (TTS) and its related time series (RTS) as the output and input of a time series estimation process, respectively. We then propose a novel data mining framework for time series estimation when TTS and RTS represent different sets of observed variables from the same dynamic system. This is made possible by mining a database of instances of TTS, its simultaneously recorded RTS, and the input/output dynamic models between them. The key mining strategy is to formulate a mapping function for each TTS-RTS pair in the database that translates a feature vector extracted from RTS to the dissimilarity between true TTS and its estimate from the dynamic model associated with the same TTS-RTS pair. At run time, a feature vector is extracted from an inquiry RTS and supplied to the mapping function associated with each TTS-RTS pair to calculate a dissimilarity measure. An optimal TTS-RTS pair is then selected by analyzing these dissimilarity measures. The associated input/output model of the selected TTS-RTS pair is then used to simulate the TTS given the inquiry RTS as an input. An exemplary implementation was built to address a biomedical problem of noninvasive intracranial pressure assessment. The performance of the proposed method was superior to that of a simple training-free approach of finding the optimal TTS-RTS pair by a conventional similarity-based search on RTS features.

Forecasting intracranial pressure elevation using pulse waveform morphology

Management of intracranial pressure (ICP) following a traumatic brain injury (TBI) is an essential aspect of minimizing such secondary brain injuries as intracranial hypertension and cerebral hypoxia. Currently, ICU management of ICP elevations is reactive in nature; we propose a quantitative method to predict potentially harmful elevations in ICP. Methods - Continuous intracranial pressure measurements were obtained from 37 patients at the UCLA Medical Center. Intracranial hypertension (IH) episodes were identified along with slow wave segments (used for control sets). Four, five minute segments were then constructed from the IH episode: one from the onset of ICP elevation (pre-IH #0) along with sets 5, 20, and 35 minutes prior to the elevation (pre-IH #5, #20, #35 respectively). Quantification and recognition of the three ICP sub peaks was performed using our group’s algorithm termed Morphological Clustering and Analysis of Intracranial Pressure (MOCAIP). Furthermore, a quadratic classifier (QDC) was used to determine the metrics with the greatest predictive power. These metrics were then used to compare the control data set to the data sets described previously. Results – From the ten most frequently selected metrics each of the four pre- intracranial hypertension (pre-IH) segments were compared with the control. Sensitivity (SEN), specificity (SPE), and accuracy (AC) were determined for each set with a SEN and SPE for the data set five minutes prior to ICP elevation of 90% and 75% respectively. Conclusion - Combining the MOCAIP analysis, QDC classification, and bootstrap method of statistical sampling, our analysis has the potential to predict an ICP elevation event 20 minutes prior to the event onset.

Improved Noninvasive Intracranial Pressure Assessment With Nonlinear Kernel Regression

The only established technique for intracranial pressure (ICP) measurement is an invasive procedure requiring surgically penetrating the skull for placing pressure sensors. However, there are many clinical scenarios where a noninvasive assessment of ICP is highly desirable. With an assumption of a linear relationship among arterial blood pressure (ABP), ICP, and flow velocity (FV) of major cerebral arteries, an approach has been previously developed to estimate ICP noninvasively, the core of which is the linear estimation of the coefficients f between ABP and ICP from the coefficients w calculated between ABP and FV. In this paper, motivated by the fact that the relationships among these three signals are so complex that simple linear models may be not adequate to depict the relationship between these two coefficients, i.e., f and w , we investigate the adoption of several nonlinear kernel regression approaches, including kernel spectral regression (KSR) and support vector machine (SVM) to improve the original linear ICP estimation approach. The ICP estimation results on a dataset consisting of 446 entries from 23 patients show that the mean ICP error by the nonlinear approaches can be reduced to below 6.0 mmHg compared to 6.7 mmHg of the original approach. The statistical test also demonstrates that the ICP error by the proposed nonlinear kernel approaches is statistically smaller than that estimated with the original linear model (p <; 0.05). The current result confirms the potential of using nonlinear regression to achieve more accurate noninvasive ICP assessment.

Improved wavelet entropy calculation with window functions and its preliminary application to study intracranial pressure

The wavelet entropy is a novel way to measure the signal regularity, and its calculation is based on the energy distribution in wavelet sub-bands. However, wavelet entropy will be largely influenced by the noise usually existed in signals, especially in physiological signals. With aim to get more stable entropy calculation, a windowed wavelet entropy approach is proposed. In this paper, we systemically studied the difference between wavelet entropy and approximate entropy, which has yet not been studied in detail before. The conducted comparison with various signals reveals that wavelet entropy can measure the signal complexity like approximate entropy. Moreover, the relative wavelet entropy can be used to measure the dissimilarity between two signals. Compared to the original wavelet entropy approach, the comparison result also shows that the proposed window approach can get smoother and more stable calculation for both wavelet entropy and relative wavelet entropy, which is more meaningful to measure signal regularity and dissimilarity. The application to the time series recorded from a patient having the intracranial hypertension reveals that the new approach can clearly differentiate the normal and hypertension states, which may serve as a promising tool for prediction of intracranial pressure in future.

Morphological changes of intracranial pressure pulses are correlated with acute dilatation of ventricles

BACKGROUNDnPotentially useful information may exist in the morphological changes in intracranial pressure pulse therefore their extraction by automated methods is highly desirable.nnnMETHODSnLong-term continuous recordings of intracranial pressure and electrocardiogram (ECG) signals were analyzed for four patients undergoing intracranial pressure (ICP) monitoring with their implanted shunts externalized and clamped. A novel clustering algorithm was invented to process hours of continuous ICP recordings such that a dominant ICP pulse was extracted every 5 min. Morphological characteristics of dominant ICP pulses were then extracted after detecting characteristics points of a dominant ICP pulse that include the locations of ICP pulse onset, the first (P1), the second (P2), and the third peaks (P3) (or inflection points in the absence of peaks).nnnFINDINGSnIt was found that ratios of P2 amplitude to P1 amplitude and P3 amplitude to P1 amplitude showed a strong increasing trend for a patient whose lateral ventricles were significantly enlarged (bi-frontal distance was 33 cm on day 0 and 37 cm on day 2) while there were no consistent trends in these morphological features of ICP pulse for the three patients whose ventricles size was not altered during the monitoring period.nnnCONCLUSIONnThe present work demonstrates the usefulness of this novel ICP pulse analysis algorithm in terms of its potential capabilities of extracting predictive pulse morphological features from long-term continuous ICP recordings that correlate with the development of ventriculomegaly.

Pulse onset detection using neighbor pulse-based signal enhancement

Detecting onsets of cardiovascular pulse wave signals is an important prerequisite for successfully conducting various analysis tasks involving the concept of pulse wave velocity. However, pulse onsets are frequently influenced by inherent noise and artifacts in signals continuously acquired in a clinical environment. The present work proposed and validated a neighbor pulse-based signal enhancement algorithm for reducing error in the detected pulse onset locations from noise-contaminated pulsatile signals. Pulse onset was proposed to be detected using the first principal component extracted from three adjacent pulses. This algorithm was evaluated using test signals constructed by mixing arterial blood pressure, cerebral blood flow velocity and intracranial pressure pulses recorded from neurosurgical patients with white noise of various levels. The results showed that the proposed pulse enhancement algorithm improved (p<0.05) pulse onset detection according to all three different onset definitions and for all three types of pulsatile signals as compared to results without using the pulse enhancement. These results suggested that the proposed algorithm could help achieve robustness in pulse onset detection and facilitate pulse wave analysis using clinical recordings.