Detection of Autonomic Dysreflexia in Individuals With Spinal Cord Injury Using Multimodal Wearable Sensors Journal: ArXiv

Bertram Fuchs, Mehdi Ejtehadi, Ana Cisnal, Jürgen Pannek, Anke Scheel-Sailer, Robert Riener, Inge Eriks-Hoogland, Diego Paez-Granados · 2025

Read Detection of Autonomic Dysreflexia in Individuals With Spinal Cord Injury Using Multimodal Wearable Sensors Journal: ArXiv by Bertram Fuchs, Mehdi Ejtehadi, Ana Cisnal, Jürgen Pannek, Anke Scheel-Sailer, Robert Riener, Inge Eriks-Hoogland, Diego Paez-Granados (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2508-03715.txt.

Introduction

Bertram Fuchs, Mehdi Ejtehadi, and colleagues detect a dangerous blood pressure emergency with wearable sensors in a 2025 medical study. Autonomic dysreflexia strikes people with spinal cord injury as sudden, severe pressure spikes that can damage the heart and brain. Current checks are invasive or rely on patients reporting symptoms, which limits daily use.

Their non-invasive framework reads several wearable signals at once and applies machine learning that can explain its alarms. Early, accurate detection is the goal, so treatment can start before complications set in. A wearable warning system could protect patients at home, far from hospital monitors.

Catalog record

CREATORBertram Fuchs, Mehdi Ejtehadi, Ana Cisnal, Jürgen Pannek, Anke Scheel-Sailer, Robert Riener, Inge Eriks-Hoogland, Diego Paez-Granados
YEAR2025
COLLECTIONPAPERS
FORMATTXT
FILE SIZE38 KB
VERSION1.0
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2508-03715

Browse the full library · About the Tower of Babel

Full text

The opening of the work, from the library's hosted text:

Detection of Autonomic Dysreflexia in Individuals With Spinal Cord Injury Using Multimodal Wearable Sensors Journal: ArXiv Bertram Fuchs, Mehdi Ejtehadi, Ana Cisnal, Jürgen Pannek, Anke Scheel-Sailer, Robert Riener, Inge Eriks-Hoogland, Diego Paez-Granados Abstract Autonomic Dysreflexia (AD) is a potentially life-threatening condition characterized by sudden, severe blood pressure (BP) spikes in individuals with spinal cord injury (SCI). Early, accurate detection is essential to prevent cardiovascular complications, yet current monitoring methods are either invasive or rely on subjective symptom reporting, limiting applicability in daily file. This study presents a non-invasive, explainable machine learning framework for detecting AD using multimodal wearable sensors. Data were collected from 27 individuals with chronic SCI during urodynamic studies, including electrocardiography (ECG), photoplethysmography (PPG), bioimpedance (BioZ), temperature, respiratory rate (RR), and heart rate (HR), across three commercial devices. Objective AD labels were derived from synchronized cuff-based BP measurements. Following signal preprocessing and feature extraction, BorutaSHAP was used for robust feature selection, and SHAP values for explainability. We trained modality- and device-specific weak learners and aggregated them using a stacked ensemble meta-model. Cross-validation was stratified by participants to ensure generalizability. HR- and ECG-derived features were identified as the most informative, particularly those capturing rhythm morphology and variability. The Nearest Centroid ensemble yielded the highest performance (Macro F1 = 0.77±0.03), significantly outperforming baseline models. Among modalities, HR achieved the highest area under the curve (AUC = 0.93), followed by ECG (0.88) and PPG (0.86). RR and temperature features contributed less to overall accuracy, consistent with missing data and low specificity. The model proved robust to sensor dropout and aligned well with clinical AD events. These results represent an important step toward personalized, real-time monitoring for individuals with SCI. 1 Introduction 1.1 Background Spinal cord injury (SCI) is a life-altering condition affecting approximately 52.5 individuals per million globally each year [ 1 ] , with over 58% of cases resulting in cervical-level injuries [ 2 ] . Beyond motor and sensory impairments, SCI frequently disrupts autonomic cardiovascular control, giving rise to autonomic dysregulation (ADys), a term encompassing both autonomic dysreflexia (AD) and orthostatic hypotension (OH) [ 3 ] . AD affects up to 90% of individuals with injuries above the T6 level [ 4 ] , characterized by rapid, often severe increases in systolic blood pressure (SBP) that can result in strokes, seizures, or even death if not treated [ 5 ] . In contrast, OH involves a significant drop in blood pressure (BP) upon standing, leading to fatigue, dizziness, and reduced quality of life [ 6 ] . Both conditions contribute to elevated cardiovascular risk in SCI individuals, especially during the early months post-injury when episodes frequently begin to manifest [ 7 ] . Prompt detection and intervention are therefore essential. Despite the clinical importance, ADys episodes often go undetected outside of hospital environments due to subtle or absent perceivable symptoms and limited awareness among caregivers [ 4 ] . Current diagnosis is primarily based on invasive monitoring or subjective patient-reported symptoms, which limits scalability and reliability in everyday settings. An objective, automated, and non-invasive detection system based on wearable sensors could enable earlier identification and timely response to ADys episodes, improving outcomes for the SCI population. 1.2 Prior Work AD remains a critical and under-monitored complication for individuals with SCI, especially those with lesions above T6. Existing studies have explored both human and animal models to develop automated AD detection systems using machine learning. Suresh et al. [ 5 ] proposed an early wearable-based system leveraging electrodermal activity (EDA), heart rate (HR), and skin temperature (Temp) sensors from a wrist-worn smartwatch. Their machine learning model (SVM) trained on self-reported AD events by SCI participants achieved high accuracy, but relied on a small sample of seven participants and lacked systematic cross-validation. In a follow-up study [ 8 ] , the same group extended their multimodal sensing approach, incorporating a more complex telemetry system and increasing the participant count to eleven. They reported a detection accuracy of 94.1%, emphasizing real-time application feasibility in community settings. Although lack of cross validation methods and small sample size, limits the generalizability of their findings. Sagastibeltza et al. [ 9 ] investigated AD onset through controlled bladder filling in a clinical environment with only five human subjects. Their approach uniquely combined physiological measurements (e.g., ECG, peripheral resistance) with hormone analysis and patient history, but it remained a preliminary proof-of-concept without wearable deployment. Pancholi et al. [ 10 ] proposed a novel approach using deep neural networks (DNNs) trained on skin nerve activity (SKNA) data collected from a rat model via colorectal distension to induce AD. Although their system achieved high accuracy (93.9% \pm 2.5%) and low false-negative rates, the use of invasive SKNA signals and non-human subjects limits the clinical translatability. Lastly, Suresh et al. [ 11 ] conducted an in-depth analysis of feature selection techniques to optimize AD detection from physiological signals. This study focused on algorithmic refinement rather than system-level deployment, further underscoring the fragmented and exploratory state of current AD detection research. These prior works highlight the potential of multimodal biosignal analysis, particularly using EDA and Temp; however, they also expose critical limitations: reliance on subjective ground truths, lack of temporal detection, and limited generalizability to everyday environments, with EDA and skin temperature being highly prone to environmental error. Currently, no study or method has demonstrated automated detection of AD episodes using non-invasive wearable sensor data, and validated it with objective BP measurements. 1.3 Goal of This Study In this study, we addressed this gap by investigating what wearable biosignals carry the most predictiveness for detecting AD. We hypothesize that these events are associated with unique physiological changes in each person, including alterations in Temp, EDA, heart rate variability (HRV), and respiratory rate (RR) [ 5 , 6 ] . Therefore, a highly non-linear relationship across signals is expected and exploited using ensemble machine learning classifiers. The contributions of this work are three-fold: (1) Identification of informative signal modalities and physiological features descriptive of AD; (2) Development of the first ensemble classifier for wearable, multi-modal AD detection based on objectively measured BP variations; (3) Design of a robust ensemble classification framework with BorutaSHAP-based feature selection, capable of maintaining performance under sensor noise or modality failure, ensuring practical applicability in real-world settings. 2 Methods 2.1 Study Population This observational proof-of-concept study included 27 individuals with chronic, motor- and sensory-complete SCI (AIS A) with level of injury (LOI) at or above the T6 level. Participants were recruited from inpatient and outpatient services at the Swiss Paraplegic Center (SPC), where they underwent a scheduled urodynamic study (UDS). Data were compiled from three ethically approved research protocols. The compiled dataset was screened for completeness, and subjects with insufficient ECG, PPG, or reference BP measurements were excluded from analysis. Demographic characteristics and sensor data availability after screening are presented in Section 3 . 2.2 Clinical Protocol and Reference AD Labels During UDS, the bladder was filled with 37°C saline solution to induce AD. SBP, DBP, and HR were recorded every 2–3 minutes using a medical-grade BP cuff. Reference AD episodes were annotated using clinical guidelines from Krassioukov et al. [ 3 ] and the American Autonomic Society [ 12 ] , which define AD as a sustained

Read the complete text