DEMS 2026 Abstracts


Area 1 - DEMS

Full Papers
Paper Nr: 5
Title:

User-Centred Design of a Cost-Efficient, Mobile, Machine Learning Based Seizure Detection Device

Authors:

Anna Katharina Hardenbicker, Gabriele Bleser-Taetz and Dorian Mora-Sanchez

Abstract: Epilepsy is a neurological condition whose main symptom, epileptic seizures, can lead to serious accidents and even death. To enable quick intervention in case of a seizure, seizure detection systems are a crucial step towards prevention of seizure-related consequences, potentially saving lives. While many of such critical seizure detection systems are meant for stationary application only, this paper presents the user-centred design and development of a mobile seizure detection system based on smart and cost-efficient technology. Requirements engineering was performed in qualitative and unstructured interviews with epilepsy patients, relatives and medical professionals to enable user-centred design. The developed wearable Internet of Things (IoT) seizure detection device was built on a Raspberry Pi Zero 2 W and an MPU-6050 accelerometer. Using a time-series random forest classification algorithm for generalized tonic-clonic seizures (GTCS) paired with SMOTE oversampling, an accuracy of 90.97% could be observed in the performance metrics. The time-critical communication of detected seizures was enabled by publishing alerts to the MQTT broker of the cloud service, where the published alerts could be sent as SMS to the caregiver. The device exhibited an average SMS timing of 50 seconds after the seizure onset with a capital expenditure (CAPEX) below 50€.
Download

Paper Nr: 6
Title:

Calibration-Enhanced Multimodal Framework for Non-Invasive Glucose Monitoring: An In-Vivo Validation

Authors:

El Arbi Belfarsi, Henry Flores, Maria Valero, Luisa Valentina Nino and Katherine Ingram

Abstract: Accurate non-invasive glucose monitoring remains a critical challenge in diabetes management. In this study, we present a multimodal framework that integrates near-infrared (NIR) spectroscopy with physiological features for non-invasive glucose prediction. The system employs a hybrid CNN architecture enhanced with Gray-Level Co-Occurrence Matrix and Fourier domain preprocessing across multiple NIR wavelengths (650, 808, and 850 nm). Validation was conducted using both glucometer readings and venous blood sample–based reference measurements to assess analytical accuracy and clinical relevance. The 650 nm wavelength consistently yielded the best performance, achieving the lowest error rates after calibration (RMSE = 16.29 mg/dL, MAE = 11.9 mg/dL, MAPE = 9.54% in glucometer-based testing). Clarke Error Grid analysis placed 89% of predictions in Zone A, while the more stringent Diabetes Technology Society (DTS) Error Grid confirmed clinical safety with 71% in Zone A and the remainder in Zone B, with no predictions in higher-risk zones. Venous sample–based validation showed greater variability but maintained the trend that calibration improved reliability, with all predictions confined to Zones A–B. These findings underscore the importance of wavelength selection and calibration in enhancing predictive accuracy and provide in-vivo evidence that multimodal, machine learning–based frameworks are feasible and clinically safe for reliable non-invasive glucose monitoring.
Download

Paper Nr: 7
Title:

OPAT@Home: A Patient-Centered Digital Tool to Support Confident IV Treatment at Home

Authors:

Harald F. Franck, Emma Milford, Jacob Justad, Beata Jahangiri, Sogeta Al-Bazi and Eunji Lee

Abstract: Outpatient parenteral antimicrobial therapy (OPAT) allows patients to receive intravenous antibiotics at home, reducing hospital burden and improving quality of life. However, OPAT poses challenges related to patient confidence, treatment adherence, and coordination with care teams. We developed OPAT@Home, an iOS application designed through a service design process to guide OPAT patients through their home-based treatment. The design stems from interviews, clinical shadowing, and expert feedback from Swedish and international stakeholders. The app includes step-by-step instructions, daily treatment tracking, symptom and photo reporting, a searchable FAQ, and direct contact with relevant care providers. Early input from experts suggests that the app might heighten clarity for patients, support communication with care teams, and lay the groundwork for safer, more individualized OPAT monitoring at home. Validation with OPAT patients is a fundamental next step to assess true usability and clinical relevance.
Download

Paper Nr: 8
Title:

Multi-Protocol Redundancy and Adaptive Sampling for Safety-Critical IoT Communication Using Clinical ECG Monitoring as Case Study

Authors:

Nina Pearl Doe and Christian Herglotz

Abstract: Reliable sensor communication is critical for Internet of Things (IoT) applications where data loss has significant consequences. This reliable data delivery is crucial in safety-critical IoT applications in domains such as healthcare, industrial automation, and autonomous systems. Systems that rely on single-protocol communications face potential limited fault tolerance and vulnerability to interference. This paper presents a multi-protocol redundant architecture that ensures 100% packet delivery through simultaneous transmission over three edge protocols (Serial (Universal Asynchronous Receiver/Transmitter (UART)), Wi-Fi, Bluetooth Low Energy (BLE)) with sequence-based deduplication, combined with dual-path backhaul (Wi-Fi and 5G). The focus of this paper is on communication-layer validation of reliability, latency, and power consumption rather than clinical efficacy assessment. The framework evaluates three themes using M5Stack ESP32-based sensor node, 2 Raspberry Pi 5 devices (one as gateway and the other as server), and 5G SA campus network. First, we evaluate edge-layer reliability through simultaneous Serial, Wi-Fi, and BLE transmission with first-arrival-wins deduplication. Second is backhaul optimization comparing Wi-Fi and fifth-generation (5G) cellular paths. Third, workload-driven adaptive sampling using high-rate electrocardiogram (ECG) streaming (250 to 1000 Hz). The architecture is validated through ECG monitoring case study with clinical-scenario-based adaptive redundancy and sampling rates. Testing across different scenarios demonstrates that multi-protocol edge transmission achieves 100% combined delivery with 2.0-2.99 times redundancy, where Serial wins 65-100% of protocol races, and Wi-Fi-BLE wireless combinations provide complementary coverage. For backhaul, Wi-Fi delivers lower average latency (2-5ms) while 5G provides better tail consistency (P99: 17-24ms versus 33-92ms), with dual-path achieving zero packet loss.
Download

Nr: 433
Title:

Objective Wearable Monitoring Identifies Activity and Health Risk Patterns in Young Adults: Implications for Monitoring System Design

Authors:

Dabin Choi and Catherine Park

Abstract: Wearable-based monitoring systems are increasingly applied in preventive healthcare to support early identification of health risks in both clinical and general populations. Young adults are typically regarded as a low-risk group due to preserved physical function and low disease prevalence and are therefore less frequently prioritized for intensive health monitoring. As a result, health assessment in this population often relies on self-reported measures. However, such approaches may not adequately capture underlying behavioral patterns, including reduced physical activity or associated psychological vulnerability. From a monitoring system design perspective, examining discrepancies between objective sensor-derived data and subjective health perception is important for evaluating current monitoring approaches and informing the development of user-centered healthcare monitoring systems. As a wearable-based health monitoring case study, 26 healthy adults aged ≥19 years (mean age ± SD: 43.15 ± 18.9 years; 9 males, 17 females) were recruited from Yonsei University Mirae Campus, South Korea, and stratified into two age groups (<40 years, n = 11; ≥40 years, n = 15). Objective physical activity was monitored for two consecutive days using a triaxial accelerometer-based wearable sensor (ActiGraph LEAP) worn on the non-dominant wrist. Outcome measures included step count, activity-related energy expenditure, moderate-to-vigorous physical activity (MVPA), sedentary time, and light-intensity activity. Subjective health perception and mental health status were assessed using validated questionnaires: the EuroQol five-dimension five-level questionnaire (EQ-5D-5L), the EuroQol visual analogue scale (EQ-VAS), DSM-5–based screening tools, the Short Geriatric Depression Scale (SGDS), and the Mini Nutritional Assessment (MNA). Between-group comparisons were performed using nonparametric statistical tests. Objective wearable-derived measures demonstrated significant age-related differences in physical activity patterns. Participants aged ≥40 years exhibited 39% higher daily step counts and 60.39% greater moderate-to-vigorous physical activity compared with younger adults, despite greater sedentary time. Younger adults showed slightly higher light-intensity activity. However, activity-related energy expenditure was only marginally higher (+4.5%), indicating lower overall physical activity levels. Subjective health perception, assessed using the EQ-VAS, was significantly lower in younger adults (78.18 vs. 91.53, p = 0.0034), whereas EQ-5D index scores did not differ significantly between groups. Elevated risks of depression, insomnia, and nutritional deficiency were observed predominantly in the younger group and were consistent with objectively measured inactivity patterns. This study indicates that wearable-based monitoring systems can identify physical activity patterns and associated health risks missed by self-reports, particularly in populations traditionally considered low-risk. The observed age-group differences should be interpreted in the context of monitoring system design rather than as evidence of age-related health advantage. From a design and evaluation perspective, integrating objective sensor-derived data with subjective assessments may support more accurate health risk identification and inform the development of scalable, user-centered healthcare monitoring systems aimed at early detection and health promotion in young adult populations.

Nr: 434
Title:

Designing a Mobile Snus Cessation Application for Young Swedish Women Using Nudge Theory and Conversational AI

Authors:

Eunji Lee and Maja Vikla

Abstract: The trend of increased snus use among young females of Swedish origin has become a pressing public health issue. Moreover, the recent influx of those using tobacco-free snus pouches has added to this serious health problem. Among young Swedish females between the ages of 16-29 years, there was a 500% increase in the prevalence of snus use over a six-year period of 2018-2024. This is a serious health crisis. Online health solutions such as mobile health apps provide new avenues. This study examines the design and development of a mobile application for snus cessation that is specifically designed for young Swedish women in the 16- to 29-year-old age group. The application integrates the concepts of nudging with the benefits that could be derived from a chatbot that uses AI to facilitate conversations. Literature review, state-of-the-art of applications incorporating digital nudges and chatbots, and a semi-structured interviews with 5 young Swedish female snus users aged 18-28 were conducted for this study. With the key findings from the aforementioned three methods, a new mobile application has been designed and developed by combining digital nudging elements, such as milestone functionality and visualizing health progress, and an empathetic conversational interface offered by a tailored chatbot. User tests reveal that the five test subjects were highly positive towards the conversational AI offering, finding that it is encouraging. A majority of the participants also felt that the nudge elements, such as the health timeline and achievements, were encouraging enough to use the new application as an integral part of their routine. This study makes an important contribution to the area of digital health by showing that the potential exists within a combination of nudge theory and conversational AI to help snus users quit. Its results show that tailored, theory-based, AI-integrated mobile health interventions can have an important part to play in dealing with new nicotine habits among young Swedish women.