Authors: Ryuzaburo Kochi (Department of Neurosurgery, Tohoku University Graduate School of Medicine, Sendai, Japan; Department of Neuroendovascular Therapy, Kohnan Hospital, Sendai, Japan; Department of Neurosurgery, Iwaki City Medical Center, Iwaki, Japan; Departments of Pediatrics and Neurology, Wayne State University, Detroit, MI), Eishi Asano (Departments of Pediatrics and Neurology, Wayne State University, Detroit, MI), Yoshiteru Shimoda (Department of Neurosurgery, Tohoku University Graduate School of Medicine, Sendai, Japan), Atsushi Kanoke (Department of Neurosurgery, Kohnan Hospital, Sendai, Japan), Shunsuke Omodaka (Department of Neuroendovascular Therapy, Kohnan Hospital, Sendai, Japan), Hiroyuki Sakata (Department of Neurosurgery, Tohoku University Graduate School of Medicine, Sendai, Japan), Kanako Sato (Department of Neurosurgery, Iwaki City Medical Center, Iwaki, Japan), Yasuhiro Suzuki (Department of Neurosurgery, Iwaki City Medical Center, Iwaki, Japan), Yasushi Matsumoto (Division of Development and Discovery of Interventional Therapy, Tohoku University Hospital, Sendai, Japan), Kuniyasu Niizuma (Department of Neurosurgery, Tohoku University Graduate School of Medicine, Sendai, Japan; Division of Development and Discovery of Interventional Therapy, Tohoku University Hospital, Sendai, Japan; Department of Neurosurgical Engineering, Graduate School of Biomedical Engineering, Tohoku University, Sendai, Japan; Department of Translational Neuroscience, Tohoku University Graduate School of Medicine, Sendai, Japan), Hidenori Endo (Department of Neurosurgery, Tohoku University Graduate School of Medicine, Sendai, Japan)
Categories: Original Research, neuroendovascular treatment, occlusion, pressure, thrombosis, time‐frequency analysis, waveform
Source: Stroke: Vascular and Interventional Neurology
Novel Use of Intracranial Arterial Pressure Waveforms to Detect Occlusive Events During Neuroendovascular Treatment
Authors: Ryuzaburo Kochi, Eishi Asano, Yoshiteru Shimoda, Atsushi Kanoke, Shunsuke Omodaka, Hiroyuki Sakata, Kanako Sato, Yasuhiro Suzuki, Yasushi Matsumoto, Kuniyasu Niizuma, Hidenori Endo
Abstract
BACKGROUND
Unexpected occlusive events, such as thrombosis, can cause serious complications during neuroendovascular treatment. Angiography provides only intermittent assessments, potentially missing rapidly developing occlusions, highlighting the need for continuous monitoring. This retrospective study assessed how accurately continuous monitoring of intracranial arterial pressure waveforms can detect occlusive events.
METHODS
We computed wavelet‐based time‐frequency amplitude of intracranial arterial pressure waveforms at frequencies ranging from 6 to 18 Hz in each of the 6444 trials obtained from 43 arteries in 37 patients. We determined whether modulation of this amplitude correctly classified the patent or occluded states defined by angiography.
RESULTS
Mixed‐model analysis revealed a significant increase in time‐frequency amplitude during occlusive events (P < 0.001). Receiver operating characteristic analysis indicated that an increase of ≥13.9% from baseline in the 6–18 Hz time‐frequency amplitude could detect occlusive events with a sensitivity of 88.5%, specificity of 87.5%, positive predictive value of 36.1%, and negative predictive value of 96.4%. A sustained increase of ≥14.3% for 5 or more consecutive trials could detect occlusive events with a sensitivity of 82.3%, specificity of 96.8%, positive predictive value of 75.5%, and negative predictive value of 94.8%.
CONCLUSIONS
Our preliminary study suggests that continuous intracranial arterial pressure monitoring holds promise as an adjunctive tool to accurately detect occlusive events. A prospective study is warranted to definitively establish its diagnostic value during neuroendovascular treatment.
Unexpected occlusive events during neuroendovascular treatment, such as thrombosis, can cause serious complications. ^1^ , ^2^ , ^3^ The incidence of thromboembolic complications during neuroendovascular treatment has been reported to range from 1% to 9%. ^1^ , ^2^ , ^3^ Early detection enables prompt intervention to minimize adverse outcomes. ^4^ , ^5^ , ^6^ , ^7^ Angiography is the current gold standard for evaluating intravascular conditions. However, it captures only brief periods during contrast injection, typically lasting a few seconds, and substantial intervals between imaging sessions remain unmonitored throughout the procedure. Providers of neuroendovascular treatment are encouraged to minimize contrast media and radiation exposure to reduce the risk of adverse events. ^8^ As a result, periods without angiographic assessment inevitably occur. Occlusions occurring during these intervals may remain undetected. Addressing these unmet needs thus requires continuous, real‐time intravascular monitoring capable of providing timely alerts precisely when angiographic confirmation of an occlusive event is necessary.
Intracranial arterial pressure (IAP) waveforms can be continuously measured via a catheter already placed for therapeutic purposes, thus requiring no additional invasive procedures. Characteristic morphological changes in IAP waveforms have been reported to reflect alterations in intravascular hemodynamics, which vary with peripheral resistance and stroke volume following arterial occlusion. ^9^ , ^10^ , ^11^ Such waveform alterations can be quantified using time‐frequency analysis. ^12^ , ^13^ Therefore, we tested the hypothesis that time‐frequency analysis of IAP waveforms can accurately differentiate occluded from patent states. Furthermore, we evaluated the predictive performance for detecting occluded states based on defined cutoff values.
METHODS
The data that support the findings of this study are available from the corresponding author upon reasonable request.
Study Inclusion
We retrospectively studied patients who underwent neuroendovascular treatment for cerebral aneurysms (ruptured and unruptured), internal carotid artery stenosis, and parent artery occlusion at Kohnan Hospital (February–March 2022) and Iwaki City Medical Center (April–December 2022). Arteries with preexisting stenosis, occlusion, or pathological vascular shunts were excluded. ^11^ , ^14^ , ^15^ The study was approved by the institutional review boards of Kohnan Hospital (2024‐1120‐07) and Iwaki City Medical Center (R6‐49) via an opt‐out procedure.
IAP Data Acquisition and Preprocessing
IAP was continuously measured using a pressure monitoring kit (Merit Medical Systems, South Jordan, UT, USA) connected to a guiding or angiographic catheter during stationary periods. Each time segment was classified as either patent or occluded based on angiographic findings. Patent states were defined as segments recorded between 2 angiographic assessments confirming patency. Occluded states included segments recorded between angiographic assessments confirming occlusion, or from the immediate period following complete balloon inflation until deflation.
IAP waveforms were filtered at 1–40 Hz and subdivided into trials corresponding to RR intervals derived from concurrently recorded ECG signals. Events containing ectopic beats were excluded. Continuous wavelet transform was applied to each IAP trial to obtain instantaneous time‐frequency amplitude values. For each trial, a specific integration window was defined (Methods S1). Within this window, time‐frequency amplitude values were integrated across 4 frequency bands (6–10 Hz, 10–14 Hz, 14–18 Hz, and 6–18 Hz). The spectral frequency band of interest was limited to 6–18 Hz because higher‐frequency components are prone to mechanical artifacts, whereas lower‐frequency components were considered unmeasurable in trials with short RR intervals. ^12^ The diagnostic utility of these integrated values was subsequently evaluated.
Mixed‐Model Analysis
A mixed‐model analysis evaluated the impact of several fixed‐effect factors on the IAP time‐frequency amplitude of each of the 4 frequency bands. These fixed‐effect factors included artery occlusion state (occlusion versus patent), patient age, diagnosis of subarachnoid hemorrhage, IAP measurement trial order within each time segment, and effective luminal area of the catheter. The effective luminal area of the outer catheter used for IAP monitoring was calculated by subtracting the total cross‐sectional area of all coaxially inserted inner catheters from the cross‐sectional area of the outer catheter. Specifically, the cross‐sectional area of the outer catheter was computed using its inner diameter, and each inner catheter's cross‐sectional area was calculated using its outer diameter. All areas were computed using the formula π·(d/2)^2^, where d represents the relevant diameter as specified by the manufacturer. Fixed‐effect variables included the diagnosis of subarachnoid hemorrhage, as it can alter arterial dynamics by increasing intracranial pressure and reducing cerebral perfusion, potentially influencing time‐frequency amplitude, ^16^ , ^17^ and the effective luminal area, as it was expected to directly affect focal arterial pressure and the IAP waveform. ^18^ , ^19^ To account for variability between arteries, both the sampled artery and intercept were included as random effects. Statistical significance was set at a Bonferroni‐corrected P value of <0.05.
Assessment of Diagnostic Performance of IAP Monitoring
Receiver operating characteristic (ROC) analysis was performed to assess how accurately the IAP‐based predictor variable could detect occlusion states as defined by angiography. The predictor variable was the percent change in time‐frequency amplitude from baseline, defined as the median value of the first 5 trials within each time segment. When the catheter's effective luminal area changed—for example, due to insertion or removal of a coaxial catheter—a new baseline was established. Diagnostic accuracy was assessed using a leave‐one‐artery‐out cross‐validation, determining optimal cutoff values by maximizing the Youden index within training sets. Classification metrics included the area under the ROC curve (AUROC), area under the precision–recall curve, sensitivity, specificity, positive predictive value (PPV), and negative predictive value (NPV). Detailed cross‐validation procedures and statistical tests are described in Methods S2.
All analyses were performed using MATLAB (R2024a, MathWorks, Natick, MA, USA; RRID: SCR_001622) and IBM SPSS Statistics (version 30, Armonk, NY, USA; RRID: SCR_002865).
RESULTS
Initially, 60 arteries from 52 patients were reviewed for eligibility. Ultimately, 43 arteries from 37 patients were included (Figure S1). The median patient age was 62 years (interquartile 50–71); 23 patients were female; 34 underwent endovascular treatment for cerebral aneurysms (ruptured in 19), 2 for neck internal carotid artery stenosis, and 1 for parent artery occlusion.
During endovascular treatment, occlusive events occurred in 14 of the 43 balloon occlusion of the parent artery (n = 9), coil occlusion of the parent artery (n = 3), and thrombotic occlusion (n = 2). A total of 124 time segments were identified, of which angiography classified 88 as patent states and 36 as occluded states. Of the 6461 RR intervals, 17 containing ectopic beats were excluded, leaving 6444 valid IAP trials (Figure S3). The median number of IAP measurement trials per time segment was 42.5 (interquartile 29.5–60.3). Further details are provided in Table 1 and Table S1.
No adverse events related to continuous IAP monitoring were noted. The Figure illustrates snapshots of angiographic findings and IAP waveforms from a patient who underwent balloon‐assisted coil embolization of an unruptured intracranial aneurysm.
Mixed‐Model Analysis
Occluded state, IAP measurement trial order, and effective luminal area were each independently associated with increased time‐frequency amplitude on IAP waveforms across all 4 frequency bands analyzed (P<0.001; Table 2); these associations remained significant after Bonferroni correction (α = 0.0125). All pairwise correlation coefficients were <0.7, indicating no substantial multicollinearity among these fixed‐effect variables. Neither age nor diagnosis of subarachnoid hemorrhage had a significant effect on time‐frequency amplitude.
Diagnostic Performance of IAP Monitoring
ROC analysis revealed that the 6–18 Hz time‐frequency amplitude provided the highest overall performance in detecting occlusive trials. Specifically, the mean Youden index averaged across 43 arteries was 0.67, with AUROC of 0.881 (P<0.001; 95% CI, 0.878–0.884), area under the precision–recall curve of 0.772 (P<0.001; 95% CI, 0.768–0.775), best cutoff of 13.9% increase (P<0.001; 95% CI, 13.8–14.1), sensitivity of 88.5% (P<0.001; 95% CI, 74.3–100), specificity of 87.5% (P<0.001; 95 %CI, 81.2–93.8), PPV of 36.1% (P = 0.09), and NPV of 96.4% (P<0.001; 95% CI, 93.0–99.8). The performance metrics of each time‐frequency amplitude predictor for individual arteries are presented in Tables S2–S5.
To assess whether brief fluctuations in time‐frequency amplitude were the primary cause of false‐positive detections in some occlusive trials, we conducted a post hoc ROC analysis. We hypothesized that a sustained increase in the 6–18 Hz time‐frequency amplitude—exceeding the cutoff for 5 consecutive trials—would improve PPV without negatively affecting other classification metrics. One time segment evaluating the left internal carotid artery in Patient 37 contained only 7 IAP measurement trials, making it impossible to assess sustained changes (≥5 trials); therefore, this time segment was excluded from the post hoc analysis.
The post hoc analysis indicated that using a sustained increase in time‐frequency amplitude reduced false‐positive detections, as evidenced by improved PPV. Specifically, the mean Youden index was 0.75, with a mean AUROC of 0.924 (P<0.001; 95% CI, 0.921–0.926), area under the precision–recall curve of 0.896 (P<0.001; 95% CI, 0.893–0.899), best cutoff of 14.3% increase (P<0.001; 95% CI, 14.2–14.4%), sensitivity of 82.3% (P<0.001; 95% CI, 64.8–99.9%), specificity of 96.8% (P<0.001; 95% CI, 92.9–100%), PPV of 75.5% (P = 0.027; 95% CI, 53.3–97.7%), and NPV of 94.8% (P<0.001; 95% CI, 90.4–99.3) (Table S6).
DISCUSSION
To our knowledge, this is the first study demonstrating that time‐frequency analysis of IAP waveforms can accurately detect occlusive events during neuroendovascular treatment. The duration of the endovascular procedure typically exceeded 2 hours across patients. Providers are generally advised to perform up to 40 angiographic imaging sessions per procedure. Even under this simplified assumption, the average interval between imaging sessions exceeds 3 minutes. Given that these intervals often span multiple minutes, transient vascular changes—such as abrupt occlusion—may go undetected. Therefore, we believe that our technique offers complementary diagnostic value by enabling continuous monitoring during clinically relevant gaps in intermittent angiography. Specifically, a sustained increase in the time‐frequency amplitude of IAP predicted occlusive trials with an AUROC of 0.924. Although these findings indicate that continuous IAP monitoring is promising for detecting occlusive events, a prospective study with a larger sample size is necessary to definitively confirm its diagnostic value. Our study is considered preliminary, as it was retrospectively conducted in a consecutive series of 37 patients who met eligibility criteria. Importantly, our results do not suggest that IAP monitoring could replace angiographic assessment; rather, they highlight its potential as an adjunctive tool to provide timely alerts precisely when angiographic confirmation of an occlusive event is needed. Given that most of the analyzed arteries were in the anterior circulation, further studies are needed to determine whether the reported diagnostic accuracy can be generalized to posterior circulation territories or adapted for use during other procedures such as flow diversion.
IAP waveforms reflect the dynamics of forward and reflected blood flows. ^9^ , ^10^ , ^11^ , ^15^ The intensity and speed of reflected blood flow have been suggested to positively correlate with the reflection coefficient, effectively representing peripheral vessel resistance. ^9^ , ^10^ , ^11^ , ^15^ Thus, the distinct notch observed on IAP waveforms, accompanied by increased time‐frequency amplitude (Figure 1B), likely reflects an earlier arrival of reflected blood flow resulting from arterial occlusion‐induced increases in vessel resistance.
The strong performance of the 6–18 Hz band may reflect differences in patient heart rates. Narrower bands may not capture such variations as effectively (Tables S2–S5). The width of the notch depended on heart rate; higher rates produced a sharper notch, making the occlusion‐related increase in time‐frequency amplitude more pronounced in the higher‐frequency band. Therefore, we plan to use the 6–18 Hz time‐frequency amplitude in the next prospective study to more definitively evaluate the diagnostic utility of continuous IAP monitoring for detecting arterial occlusive events.
The results of our mixed‐model analysis indicate that capturing the relative increase in IAP time‐frequency amplitude over time may be more important than absolute amplitude values. Reductions in the effective luminal area significantly decreased the time‐frequency amplitude. This suggests that inserting additional coaxial catheters diminishes waveform magnitude. ^18^ , ^19^ Therefore, when changes in the effective luminal area occur, recalibrating the baseline IAP time‐frequency amplitude is recommended to optimally detect occlusive events.
We found no evidence suggesting IAP monitoring is challenging in patients with subarachnoid hemorrhage. No adverse events were observed in any of the study patients. Our mixed‐model analysis found no significant effect of subarachnoid hemorrhage diagnosis on IAP time‐frequency amplitude. Although elevated intracranial pressure reportedly influences cerebral perfusion, ^16^ , ^17^ we successfully performed continuous IAP monitoring and detected increases in time‐frequency amplitude associated with transient arterial occlusion (Table S6).
Later trials were associated with increased time‐frequency amplitude within the given time segments, although the effect was modest (0.003 mmHg·s per trial number in 6–18 Hz band; Table 2). One possible explanation is slowly developing vasospasm caused by endovascular manipulations, such as adjustments of catheters or balloons. Nonetheless, this trial effect may be clinically negligible, given that the magnitude of time‐frequency amplitude increase associated with arterial occlusion was 3.019 mmHg·s.
The reported low PPV (36.1%) in the per‐trial analysis is primarily due to false positive detections unrelated to true occlusive events. These false positives likely result from frequent, intrinsic fluctuations in the IAP waveform, which can produce transient time‐frequency amplitude augmentations that mimic true occlusive events. We also evaluated the PPV using a more stringent detection an event was classified as occlusive only when the time‐frequency amplitude augmentation persisted across at least 5 consecutive trials. This stricter threshold substantially reduced the false positive rate and improved the PPV to 75.5%, while modestly reducing the sensitivity from 88.5% to 82.3%. Furthermore, the study design, in which 29 of the 43 study arteries did not exhibit true occlusive events, also contributed to the relatively low PPV. Even when using diagnostic tools with high sensitivity and specificity, including a substantial number of patients or assessments without true positive events can reduce the PPV, as false positives comprise a larger proportion of the positive detections. ^20^
The validity of angiographic examinations as the gold standard for defining arterial occlusion events warrants discussion. Angiography provides intermittent assessments; thus, we had to assume that the intravascular state (occluded or patent) remained stable between consecutive angiographic examinations. Angiography may fail to detect transient thrombosis‐related occlusions occurring between evaluations or transient recanalization immediately following documented occlusions. Figure S3 illustrates dynamic modulations of IAP time‐frequency amplitude in Patient 01, who experienced a thrombotic event during neuroendovascular treatment. This patient was 1 of 2 patients in whom the accuracy of IAP‐based occlusion detection was suboptimal. Figures S3C, S3J, and S3K demonstrate fluctuations in IAP time‐frequency amplitude even during angiographically defined occlusion states. We cannot rule out arterial instability due to fluctuating thrombotic conditions. After recognizing this occlusive event on angiography, fasudil hydrochloride and ozagrel sodium were administered. Subsequent angiograms revealed alternating patent and occluded states, supporting our suspicion of rapid intravascular changes. Eventually, the thrombus resolved completely without neurological deficits. This anecdotal observation led us to hypothesize that continuous IAP time‐frequency analysis could serve as an adjunctive tool, providing real‐time insights into rapidly changing intravascular conditions.
Sources of Funding
This study was funded by JSPS KAKENHI Grants 19K18376 and 25K19895.
Disclosures
None.