Deep learning-based arterial waveform analysis for predicting postoperative cerebrovascular events in pediatric
Jung-Bin Park1, Youmin Shin2,3, Jihun Kim2,4
1Department of Anesthesiology and Pain Medicine, Seoul National University Hospital, College of Medicine, Seoul National University, Republic of Korea.
Insights
Deep learning models analyzing intraoperative arterial blood pressure (ABP) waveforms can predict postoperative cerebrovascular events in pediatric Moyamoya disease (MMD) patients. Diastolic runoff dynamics show potential as a key indicator for these events.
Area of Science:
- Neuroscience
- Biomedical Engineering
- Artificial Intelligence in Medicine
Background:
- Postoperative cerebrovascular events pose a significant risk for pediatric patients with Moyamoya disease (MMD) undergoing revascularization surgery.
- Predicting these events is crucial for improving patient outcomes and surgical strategies.
Purpose of the Study:
- To develop an explainable deep learning model for predicting postoperative cerebrovascular events in pediatric MMD patients.
- To utilize intraoperative arterial blood pressure (ABP) waveform analysis for this prediction.
- To explore waveform-derived physiologic features associated with these events.
Main Methods:
- Retrospective analysis of 181 pediatric MMD patients, with validation on a separate cohort of 79 patients.
- Preprocessing of ABP signals and conversion into image representations for deep learning.
- Evaluation of Convolutional Neural Network (CNN) and Vision Transformer (ViT) models, including ResNet and VGG architectures.
- Utilized Grad-CAM for visualization and analyzed waveform-derived features.
Main Results:
- CNN models, particularly using raw pulse waveforms, outperformed ViT models, achieving an AUROC of 0.772 internally and 0.738 in the validation cohort.
- Grad-CAM analysis identified the diastolic runoff phase as important for classification.
- Four features related to arterial compliance were significantly associated with postoperative events (p < 0.05).
Conclusions:
- Convolutional Neural Network (CNN)-based deep learning models show feasibility in predicting postoperative cerebrovascular events from intraoperative ABP waveforms.
- Diastolic runoff dynamics in ABP waveforms may represent a relevant physiologic pattern for event prediction.
- Findings are exploratory and necessitate prospective, multi-center validation for clinical application.
Background:
Postoperative cerebrovascular events, including transient ischemic attacks, infarctions, and hemorrhages, remain a significant concern in pediatric patients with Moyamoya disease (MMD)undergoing surgical revascularization. This study aimed to develop an explainable deep learning-based classification model using intraoperative arterial blood pressure (ABP) waveform analysis for postoperative cerebrovascular events in pediatric patients undergoing surgery for MMD, with exploratory analysis of associated waveform-derived physiologic features.
Methods:
This retrospective study included 181 pediatric patients (≤18 years) who underwent revascularization surgery for MMD, with an independent temporal holdout cohort of 79 patients reserved for validation. ABP signals were preprocessed using detrending, pulse segmentation, and normalization, then converted into image representations for deep learning classification. Various convolutional neural network (CNN) models, including ResNet50, ResNet34, DenseNet121, VGG16, and VGG19, were evaluated against Vision Transformer (ViT) architectures. Multiple image transformation methods were tested, and Grad-CAM analysis and statistical comparisons of waveform-derived physiologic features were conducted between patients with and without postoperative cerebrovascular events.
Results:
The optimal model configuration achieved the best performance using raw pulse waveforms with three consecutive pulses per image. CNN-based models outperformed ViT-based models, with the highest internal classification performance observed using raw pulse waveforms (AUROC = 0.772, SD = 0.070).In the independent temporal validation cohort, the model achieved an AUROC of 0.738 ± 0.011 at the patient level. Grad-CAM visualization highlighted the diastolic runoff phase as a region of interest for classification. Four waveform-derived features related to arterial compliance were significantly associated with postoperative cerebrovascular events (p < 0.05).
Conclusions:
In this study, CNN-based deep learning models demonstrated the feasibility of predicting postoperative cerebrovascular events from intraoperative ABP waveforms, with diastolic runoff dynamics emerging as a potentially relevant physiologic pattern. These findings are exploratory and require prospective multi-center validation before clinical application.
