BIODEVICES 2026 Abstracts

Full Papers Short Papers
Full Papers
Paper Nr: 157
Title:

Seismocardiography (SCG) Signal Classification of Valvular Heart Diseases with Convolutional Neural Network
Authors:

Mahsa Raeiati Banadkooki and Martin Bogdan
Abstract: Valvular heart disease (VHD) implies the impairment of the functioning of the heart's valves, the mitral, tricuspid, aortic, and pulmonary valves. In this study, Continuous Wavelet Transform (CWT) images were obtained from Seismocardiography (SCG) signals of VHD patients, and a multiclass convolutional neural network (CNN) model was developed for multilabel classification of four types of VHD. The dataset included records obtained from 100 patients and whose types of VHD were varied. The CNN model was satisfactory in classifying the different types of VHD. The proposed model was able to achieve a macro-averaged F1-Score of 0.66 and a Weighted-average F1 of 0.68. This work provides insights regarding the utilization of SCG signals and CNN model based on CWT for the diagnosis of VHD disease in a non-invasive and precise manner.
Paper Nr: 158
Title:

Prototype UV-A Source for Evaluating Sunglass Frame Efficacy in Blocking Environmental Radiation
Authors:

Pedro Guedes and Liliane Ventura
Abstract: This study seeks to assess the efficacy of sunglass frames in shielding the end user's eyes from ambient diffuse light. While existing standards for sunglasses primarily concentrate on defining safe limits for lens transmittance within the 280 nm to 380 nm range, it is crucial to acknowledge the role of sunglass frames in protecting the end user's eyes from ultraviolet exposure. To accomplish this objective, a prototype has been devised to replicate diffuse light environments and gauge the frame's effectiveness in blocking the light reaching the eyes when wearing sunglasses. The prototype consists of an LED-illuminated sphere, a mannequin head equipped with optical sensors and a comprehensive set of baseline and measurement procedures. Tests were conducted at the Laboratory of Ophthalmic Instrumentation (LIO/EESC/USP) unveiled variations in the light-blocking capabilities among different frame types, demonstrating a light-blocking efficiencies ranging from 65.19% to 97.57%.
Paper Nr: 179
Title:

Impedance Spectroscopy-Based Monitoring of Cellular Adhesion and Stimulus Response on a Sensorized Multi-Well Plate
Authors:

Ludovica Pradetto Battel, Lara Franchin, Alberto Zuanon, Girolamo Calò, Davide Malfacini, Alessandro Paccagnella and Stefano Bonaldo
Abstract: Real-time monitoring of cellular processes is critical for advancing quantitative cell assays. This work investigates the influence of cell adhesion on the electrical response of electrodes in a sensorized multi-well plate, aiming to establish a versatile platform for real-time cell assays employing Electrochemical Impedance Spectroscopy (EIS) as a non-invasive technique to monitor cell dynamics. The objective of this work is to develop a versatile analytical framework capable of modeling the impedance response into a lumped electric circuit, where the distinct electrical elements may provide deeper insight into the concurrent biological processes occurring at the cell/electrode interface. A low-cost, portable measurement setup is implemented by integrating a potentiostat with an Arduino-based well-selection system. CHO cell adhesion is tracked over 210 minutes of observation time, and the resulting impedance data are modeled using equivalent electrical circuits. Results show that cell adhesion increases resistance of 117%, due to hindered charge transfer, and decreases capacitance of 72%, reflecting progressive cell coverage on electrodes. After 3.5 hours, cells expressing kappa-opioid receptors are stimulated with Dynorphin A, producing significant variations in electrical parameters, featuring a sharp transient within the initial 10 minutes, with resistance peaking at 110% and capacitance dropping by 160%. These results represent a proof of principle for the feasibility of performing real-time, quantitative cell assays through impedance-based monitoring, and to attempt response deconvolution.
Paper Nr: 310
Title:

Ring-Magnet-Based Module for Efficient Magnetic Bead Mixing in a Closed Digital Microfluidic Chamber
Authors:

Chuan Lyu, Huaqing Zhang, Yuping Zhu, Jing Wang, Xuesong Ye and Congcong Zhou
Abstract: Magnetic digital microfluidics (MDMF) offers significant advantages in biological analysis and detection, including high flexibility, high efficiency, and low sample consumption. As a key functional component of MDMF, magnetic beads require efficient mixing within droplets, which directly affects the accuracy and reliability of detection results. In this work, we propose a novel magnetic bead-based mixing strategy tailored for MDMF systems. By developing a fixed ring magnet in combination with repeated chip vibrations, efficient and stable mixing of magnetic beads was achieved within a water-in-oil droplet confined in a closed chamber. Through simulations and experiments, we demonstrated that the ring-magnet-based mixing module provided more uniform bead distribution than other mixing techniques, enabling excellent cleaning effect under various assay conditions. This module provides a simple, efficient, and controllable magnetic bead mixing method that can positively influence the overall performance of immunoassays and is suitable for system integration, offering a practical solution for the modular design of magnetic digital microfluidics.
Paper Nr: 319
Title:

A Validated Computational Workflow for 2D PTV: From Images to Eulerian Fields for Biofluid Analysis
Authors:

Raissa Ré, Andrews Souza, Rui Lima, Maurício Ferreira, Carlos E. Baccin and Ana S. Moita
Abstract: In vitro studies are becoming an important approach for replicating and understanding the complex flow characteristics that may inform the diagnosis and monitoring of cardiovascular diseases. In this context, the quantification of haemodynamic parameters in vascular flows has been widely investigated using advanced particle tracking techniques. However, procedures to obtain reliable processed data, such as velocity fields from raw images, in such complex flows is still very challenging. This paper presents a complete, validated computational workflow for 2D PTV in patient-specific vascular models. The pipeline integrates image preprocessing, particle identification and tracking, and a MATLAB workflow featuring binning, universal outlier detection and interpolation. The methodology was validated through three tests: (1) accuracy against synthetic ground-truth data (Mean Relative Error < 4.3%), (2) feasibility on an idealized in vitro vessel section, and (3) robustness via a repeatability study on a complex aneurysm model (Coefficient of Variation < 5%). The results demonstrate that this workflow is a reliable, accessible tool for transforming raw images into high-fidelity velocity fields for biofluid research.
Paper Nr: 396
Title:

Mapping of Electrode Positions in the Capture of Myoelectric Signals from the Cervical Region to Optimize the Application of Machine Learning
Authors:

Vitor Ferreira Paschoal, Elizangela Almeida de Carvalho and Alberto Cliquet Jr
Abstract: This work presents a robust, high-performance framework for classifying electromyographic (EMG) signals from cervical muscle activity, enabling advanced assistive technologies for individuals with tetraplegia. Our methodology uses a strategic electrode configuration targeting the sternocleidomastoid (SCM) and splenius capitis muscles to capture distinct activation patterns for head movement. A rigorous signal processing pipeline incorporating temporal normalization, filtering, and wavelet denoising was used to compute discriminative time-domain features. The machine learning investigation employed a suite of deep learning architectures, including CNNs, LSTMs, and Transformers. Experimental results demonstrate exceptional performance, with a 1D-CNN model achieving 89.74% accuracy using Waveform Length features. Our systematic ensemble strategies consistently surpassed individual models, with a feature-specific ensemble matching the top performance. This study delivers a complete, high-accuracy framework for translating neck muscle activity into reliable command signals, providing a viable solution for creating non-invasive, EMG-driven control interfaces to restore autonomy for individuals with tetraplegia.
Paper Nr: 399
Title:

Non-Invasive Detection of Diabetes: An SVM and Wavelet-Based Approach Using the CBmeter Methodology
Authors:

Clara Ribeiro Neves, Rafael Fernandes Pinheiro, Maria Pedro Guarino and Rui Fonseca-Pinto
Abstract: Diabetes mellitus is a common chronic metabolic disorder with significant global health and economic impacts. Early detection and management are essential to reduce complications as the number of diabetes cases is expected to increase. This study investigates the use of machine learning models, specifically Support Vector Machines (SVM), enhanced by wavelet transformations for early diabetes detection. The data set includes physiological signals such as heart rate (HR), respiratory rate (RR), and oxygen saturation (SpO₂), recorded from 33 participants in a multicenter study. Various SVM kernels, including Linear, Polynomial, and Radial Basis Function (RBF), were evaluated, with and without wavelet transformation, to assess their predictive performance. Preliminary results suggest that the Linear kernel combined with three-level wavelet decomposition may offer improved performance, yielding an accuracy of 94.0%, precision of 95.8%, and recall of 92.0% in this experimental setting. In contrast, using summary statistics (mean and variance) without wavelet transformation, the Linear kernel achieved 87.88% accuracy and 100% precision but demonstrated significantly lower recall (66.67%). While wavelet transformations successfully enhanced sensitivity and overall accuracy for the Linear model, the performance of the RBF kernel declined with wavelet integration. Despite the limitations of a small sample size, these findings indicate that incorporating wavelet-based features has the potential to improve model sensitivity, a critical metric for non-invasive early diabetes detection.
Paper Nr: 402
Title:

From Person-Dependent to Person-Independent EMG-Based Action Unit Recognition of Subtle and Intense Expressions
Authors:

Dennis Küster, Rathi Adarshi Rammohan, Yvo Muskulus, Romina Razeghi Oskouei, Eva G. Krumhuber, Tanja Schultz and Rainer Koschke
Abstract: Facial expressions are at the center of everyday social interaction and communication. However, they are often degraded or absent in mediated settings such as virtual reality or as a result of neurological and motor disorders. Electromyography (EMG) offers a sensitive and privacy-preserving alternative to camera-based approaches for detecting facial Action Units (AUs). Yet although the feasibility of EMG-based Action Unit Recognition (AUR) has been demonstrated, prior work has largely focused on peak-intensity expressions, limited AU sets, and single-trial recordings. To address this gap, the present study recorded 10-trial EMG at six facial muscle sites (three upper face, three lower face) from four subjects who produced nine AUs (including neutral) at two intensity levels (subtle, peak-intensity), resulting in a 17-class classification task. We further compared the performance of two types of mobile electrodes with the biosignalsplux system. The best results for all classes yielded 83% accuracy (Random Forest) for person-dependent models and 39% accuracy for person-independent models (Gaussian Naive Bayes). Increasing the amount of training data yielded only modest gains for person-independent models, whereas person-dependent models achieved robust performance, even under few-shot learning conditions with shallow machine learning models. In contrast to previous single-trial findings, larger snap-on electrodes consistently outperformed smaller reusable electrodes, suggesting increased robustness during repeated and high-intensity activations. We discuss how an end-to-end pipeline that maps EMG-based AUR outputs to FACS-accurate animations of photorealistic virtual humans could further advance EMG-based AUR as a viable privacy-preserving approach for future interactive and assistive applications.
Short Papers
Paper Nr: 64
Title:

Microfabricated in Vitro Platform for Characterizing Cellular Calcium Dynamics During Ultrasound Neuromodulation Using Calcium Indicators
Authors:

Runo Kitahara and Takashi Tateno
Abstract: Ultrasound stimulation (US) is a minimally invasive neuromodulation technique that is capable of targeting deep brain regions with high spatial resolution. However, the neuronal response mechanisms to US at the cellular level remain unclear. Elucidating these mechanisms in brain slices—in which the maintenance of circuit-level connectivity is essential—remains an important challenge. Here, we developed an in vitro experimental system to enable the high-fidelity detection of Ca2+ dynamics during US of the mouse auditory cortex. First, Ca2+ imaging using Fluo-4 AM revealed heterogeneous Ca2+ response patterns, including excitatory, inhibitory, and biphasic dynamics across cellular populations. To improve detection accuracy, we developed a dual-wavelength excitation system in slices expressing the genetically encoded Ca2+ indicator GCaMP and performed pseudo-ratiometric imaging and analysis. This approach reduced photobleaching artifacts and enabled a more reliable extraction of genuine Ca2+ dynamics in neural populations. Our results demonstrate the utility of this system for quantifying US-driven neuronal activity in brain slices. Our system thus provides a robust platform for investigating the cellular mechanisms underlying US neuromodulation.
Paper Nr: 156
Title:

Automated Multilabel Classification of Valvular Heart Diseases Using Aortic Valve Opening Variability Features and Wavelet Scattering Transform
Authors:

Mahsa Raeiati Banadkooki and Martin Bogdan
Abstract: We propose an automated multilabel classification model for Valvular Heart Disease (VHD) using Seismocardiogram (SCG) signals. Aortic Valve Opening (AOV) variability features with classical models achieved modest performance. To overcome this, we applied Wavelet Scattering Transform (WST) to raw SCG signals and Continuous Wavelet Transform (CWT) images, followed by Machine Learning classification. WST applied to CWT images significantly outperformed all other approaches. This shows that time–frequency representations based on scattering can be useful in VHD multilabel classification. The aim of this study is to evaluate whether scattering-based time–frequency representations of SCG signals can improve multilabel classification of valvular heart diseases, achieving a weighted F1-score of 0.60 with the proposed CWT–WST pipeline.
Paper Nr: 188
Title:

Evaluation of Charge Injection Based on Electrode Model in Stem-Cell Electric Stimulation Assays
Authors:

A. Algarín, D. Martín, P. Daza, G. Huertas and A. Yúfera
Abstract: This work progress towards the electrode design process for Electric-Stimulation (ES) assays of stem-cells developed in tissue engineering. The selection of the optimal issues both for biological material (cells) and circuits involved, can be strongly dependent on electric performance of electrodes. We will be described some issues, that conclude that injected charge driven by the electrodes can be the best parameter for selecting optimal ES designs in stem cell differentiation processes. Based on our assays in ES with Neuroblatom (N2a) cells, the optimum conditions for ES said us that an injected charge density below 1.0 µC/cm2 by period, defines the optimum conditions to promote N2a differentiation toward neurons.
Paper Nr: 189
Title:

Wearable sEMG Monitoring System toward Axial Symptoms Assessment in Parkinson’s Disease: Preliminary Study
Authors:

Michele A. Gazzanti Pugliese di Cotrone, Franco Capone, Martina Patera, Silvia Gallo, Antonio Suppa, Carlo Alberto Artusi, Gabriele Imbalzano and Fernanda Irrera
Abstract: Monitoring axial impairment in Parkinson’s disease (PD) requires reliable and objective tools that go beyond traditional clinical evaluation. This study presents a wearable monitoring device based on surface EMG for assessing the severity of axial motor symptoms during a standing task. Thirty-one PD participants performed a 60 s eyes open standing task while bilateral sEMG was acquired from thoracic and lumbar back muscles. Signals were filtered, and a 40 s segment was analysed. Twelve time and frequency domain features were extracted, normalized and converted into asymmetry indices for each right–left pair. Principal Component Analysis (PCA) reduced data dimensionality, and Canonical Correlation Analysis (CCA) linearly combined the PCA components with commonly used clinical scales to derive an integrated biomarker of axial symptom severity. Empirical correlations obtained through 5,000 permutations showed stronger and more stable associations than analytical estimates. The proposed monitoring device provides a quantitative framework that minimizes operator variability and merges multiple clinical dimensions into a single objective index, supporting continuous and standardized monitoring of axial symptoms in Parkinson’s disease.
Paper Nr: 190
Title:

Electrical Stimulation Bystander Effect Promotes Neuronal Differentiation and Neurite Outgrowth: Preliminary Results
Authors:

D. Martin, L. Portillo, A. Algarín, A. Yúfera, N. Pastor, L. Orta and P. Daza
Abstract: This study investigates the bystander effect associated with the electrical stimulation (ES) of mouse neuroblastoma cells (N2a). While the direct stimulation of N2a cells has been previously characterized and shown to promote neuronal differentiation, the potential for ES to induce similar effects in non-stimulated neighboring cells remains unclear. Here, we demonstrate that the bystander effect can mimic direct ES, leading to enhanced neuronal differentiation and neurite outgrowth in non-stimulated cells. These findings suggest that ES may influence larger cell populations than previously thought, which could have significant implications for neural tissue engineering and regenerative medicine.
Paper Nr: 194
Title:

Preliminary Work on the Stratification of Myocardial Ischemia Using a Deep Learning Algorithm on 12-Lead Electrocardiograms
Authors:

Adam Butchy, Michael Leasure, Utkars Jain, Veronica A. Covalesky and Gary S. Mintz
Abstract: Each year, more than 7 million Americans report to hospitals with chest pain. Determining the origin and severity of chest pain can be a diagnostic challenge, with early stage testing lacking accuracy in recognizing cardiac ischemia. In this study we design, train, and validate a novel deep learning model to detect myocardial ischemia from electrocardiograms (ECG). To train our model, we use the THEW database, consisting of 927 patients and their ECGs before, during, and after single-photon emission computerized tomography (SPECT) stress testing. In classifying a patient between healthy and ischemic, the F-1 Score was 0.991; while in classifying a patient between low and high levels of ischemia, the F-1 Score was 0.984. This preliminary work demonstrates the feasibility of an AI model to detect different levels of myocardial ischemia from an ECG, and suggests that this framework can possibly be used to detect myocardial ischemia on chest pain patients in the emergency room.
Paper Nr: 228
Title:

Wireless Power Transfer for Neonatal ECG Monitoring: Technologies, Challenges, and Future Prospects
Authors:

Inas El-Aroussi, Purav Shah, Andreas Demosthenous and Richard Bayford
Abstract: Continuous monitoring of neonatal electrocardiogram (ECG) is an essential aspect of diagnostics in neonatal intensive care units (NICUs). Traditional wired systems restrict mobility, obstruct skin-to-skin bonding, and increase infection risks, while battery-powered alternatives are constrained by limited lifetime and thermal safety issues. Wireless power transfer (WPT) offers a promising solution for unobtrusive, contactless, and continuous energy delivery to neonatal monitoring devices. This paper provides an overview of WPT fundamentals, current state-of-the-art for neonatal ECG monitoring, challenges inherent for implementing WPT and considerations for real WPT implementation within NICU environments. A particular focus is placed on the critical role of power transfer efficiency (PTE), which is essential for enabling multi-sensor powering and battery-assisted operation to support skin-to-skin bonding also known as kangaroo care.
Paper Nr: 287
Title:

Optimizing Channel Geometry for Passive Mixing in Thermal Transfer Printed Paper-Based Microfluidic Devices
Authors:

Aujchara Thepbut, Peeraphan Compiro, Pornchai Keawsapsak and Pimkhuan Hannanta-anan
Abstract: Paper-based microfluidic devices offer a promising platform for low-cost and portable bioassays. However, effective mixing of reagents within microscale channels remains a major challenge. In this study, five passive mixing geometries-T-shape, Y-shape, dashed-shape, hourglass-shape, and cross-shape-were fabricated using thermal transfer printing and systematically evaluated based on their mixing performance using two colored solutions. Among the tested designs, the cross-shape microfluidic achieved the highest mixing efficiency (56.77%), followed by the Y-shape and T-shape configurations. In contrast, the dashed and hourglass designs showed inconsistent performance due to limitations in printing resolution. To demonstrate practical utility, the optimized cross-shape microfluidic was integrated with a CRISPR– Cas12a assay for the detection of Mycobacterium tuberculosis (TB). The device effectively mixed the test sample with the CRISPR reagents and generated clearly distinguishable fluorescence signals between positive (TB DNA template) and negative (DEPC-treated water) samples. Overall, this work highlights the importance of mixing optimization in paper-based microfluidics and establishes thermal transfer printing as a rapid and accessible fabrication method for developing practical paper-based nucleic acid diagnostics. The findings provide valuable insights for advancing paper-based microfluidic design and expanding their applications in molecular diagnostics and point-of-care testing.
Paper Nr: 335
Title:

Accelerometer-Based MET Estimation for Wearable Energy Expenditure Monitoring
Authors:

Adam Neulander, Dayyan Chaudhri, Joseph Nash, Nathan Yasnovsky and Delaram Yazdansepas
Abstract: Accurately estimating energy expenditure (EE) from wearable devices remains challenging without heart-rate sensing or indirect calorimetry. This work investigates a lightweight, accelerometer-based framework for estimating metabolic equivalent (MET) intensity levels using only wrist- and hip-mounted triaxial accelerometers. We extract ENMO-derived features and apply published nonlinear mappings to estimate MET values without user calibration, multi-sensor fusion, or physiological measurements. To assess the plausibility of these estimates in the absence of calorimetric ground truth, we introduce an indirect validation strategy based on three criteria: (1) consistency with benchmark MET tables, (2) cross-subject stability, and (3) correct intensity ordering across walking, jogging, and stair-related activities. Results show that accelerometer-only MET estimation captures expected physiological patterns and separates activity types in a realistic and interpretable manner. These findings highlight the potential of ENMO-based methods for real-time wearable EE monitoring, particularly in applications where relative intensity tracking is more important than precise caloric accuracy.
Paper Nr: 393
Title:

Investigating the Relationship of Functional Electrical Stimulation Signal Parameters to Finger Flexion and Discomfort
Authors:

Danica Marie A. Dumalagan, Klyle Alexandre T. Luchavez, Jun Niel T. Paquibot, Clyde Matthew Y. Condor and Luis Gerardo S. Cañete Jr.
Abstract: Functional Electrical Stimulation (FES) induces functional muscle contraction by delivering electrical signals to motor nerves through electrodes. Existing FES systems use a wide range of stimulation parameters, even when targeting the same limbs and functions, highlighting the need for further investigation. This paper presents an analysis of FES stimulation signal parameters for finger flexion. The influence of amplitude, shape, burst frequency, and carrier frequency on both finger flexion and user-perceived discomfort is examined.
Paper Nr: 413
Title:

Development and Integration of a Functional Electrical Stimulation System for Selective Finger Extension
Authors:

Andre Kimberly Trish D. D'Silva and Luis Gerardo S. Cañete Jr.
Abstract: Functional Electrical Stimulation (FES) has been widely explored for restoring hand function following stroke, with most systems prioritizing finger flexion to enable grasping. However, many individuals with upper-limb spasticity already exhibit involuntary finger flexion, making reliable finger extension the primary unmet requirement for functional hand opening. This paper reframes finger extension as the dominant clinical and technical bottleneck in FES-based hand rehabilitation. We present an automated, extension-oriented FES platform that combines high-density electrode mapping with a compact, scalable solid-state switching architecture. A 36-electrode matrix is deployed on the dorsal forearm to increase spatial resolution over anatomically dense extensor motor points. Deterministic electrode exploration is performed without learning-based optimization, and finger-specific joint angle responses are quantified using an IMU-based feedback system. Experimental results with healthy participants show that finger extension occurs in narrow, spatially localized regions that are frequently missed by sparse electrode layouts. Compared to a lower-density configuration, the proposed matrix enables more consistent and selective extension mapping across fingers. The scalable switching architecture preserves stimulation fidelity while enabling practical expansion to higher electrode counts. These findings demonstrate that high electrode density and hardware scalability are prerequisite enablers for reliable finger extension in FES.
Paper Nr: 66
Title:

Embedded PPG in a Smartphone Case for non-Invasive Blood Pressure Estimation
Authors:

Alberto Alfonso García, Ricardo Del Rio García and Juan A. González Sánchez
Abstract: We propose integrating a PPG sensor directly into a smartphone cover to measure blood pressure in a single step: simply place a finger on the sensor. The system connects Bluetooth to a mobile app that automatically records measurements, displays results as graphs and tables, and enables easy tracking and sharing with healthcare providers. The study also includes signal-processing elements and the implementation of algorithms aimed at optimizing measurement accuracy and reducing noise and motion artifacts. This novel, everyday solution simplifies daily monitoring and can help prevent complications by detecting elevated blood pressure early and saving lives.
Paper Nr: 175
Title:

Visual Field Light Blocking of Sunglass Frames: A Quantitative Imaging Method
Authors:

Pedro T. X. da Silva and Liliane Ventura
Abstract: This work proposes an imaging methodology to quantify the light-blocking performance of sunglass frames. This is an additional safety parameter not considered by current standards, which directly impacts the peripheral light that reaches the eye, and it allows understanding how light blocking varies as a function of frame geometry. The method uses an experimental apparatus with a headform and a software pipeline to segment images into weighted Frontal (81.3%) and Lateral (18.7%) Regions of Interest (ROIs). These ROIs are mathematically modelled on the standard human field of view. The frame is identified via threshold-based binarization, and the blocking percentage is calculated. To validate the method, 5 samples with different geometries were tested. The results showed low mean standard deviation between eyes (0.4%), and significant difference between samples (a 6.0% range). This methodology provides a new metric for characterizing the protective contribution of sunglass frame geometry.
Paper Nr: 192
Title:

Method for Obtaining the Relative Proportions of Ultraviolet Radiation Incident on the Eyes as a Function of Different Sunglasses Frame Geometries Using Eletronic Instruments
Authors:

Augusto Perez de Andrade and Liliane Ventura
Abstract: This study aims to evaluate the percentage of ultraviolet radiation blockage achieved by using different sunglasses frame geometries. To this end, an eletronic prototype using LEDs and photodiodes was developed in a controlled spherical environment with a mannequin head. The test was able quantify the portions of radiation transmitted by the lens, the radiation reflected by the inner surface of the lens, and the peripheral radiation incident on the sides of the frame. The results obtained made it possible to associate sunglasses models with frames that allow the incidence of up to 12% of diffuse radiation from the environment on the eyes.
Paper Nr: 408
Title:

Strategies for Low-Cost, High-Quality Artificial Intelligence and Internet of Things Interventions to Promote Women’s Health
Authors:

Narasimha Sai Yamanoor and Srihari Yamanoor
Abstract: Women constitute a major component of medically underserved and unserved cohorts across many classifications, especially in developing and underdeveloped nations. Public Health interventions designed to assist women must be developed with safety and effectiveness in mind, while maximizing reach, simplicity, affordability, and ease of use. In addition, cost is an important factor, and efforts are required to remove it as a barrier to implementation, adoption, and adherence. Technological innovations in Artificial Intelligence (AI) and Internet of Things (IoT), alongside smartphones and communication modalities such as Bluetooth Low Energy (BLE), can assist intervention designers in accomplishing these goals with a high degree of success. Strategies that optimize both the technical and the softer aspects of solution design are presented, along with examples from past and ongoing work.
Paper Nr: 436
Title:

Development of an In-Ear Wearable Device for Continuous Monitoring of Elderly Patients with Dementia: A Preliminary Approach
Authors:

José Soares, Ana Rita Freitas, Marcelo Arantes, Inês Rocha, Mariana Carvalho, Marta Pinto, Demétrio Matos, Pedro Morais and Vítor Carvalho
Abstract: The global increase in the elderly population presents significant challenges in the management of age-related neurodegenerative diseases, such as dementia, placing growing demands on continuous health monitoring and support for healthcare caregivers. This paper presents the development of a comprehensive wearable system designed to support the physiological monitoring and care of elderly individuals with dementia. The system integrates two devices: the HowMi in-ear device and a wristband, enabling the continuous and non-invasive acquisition of vital signs and activity-related data. This study focuses on the implementation of an initial prototype of the in-ear device, designed to monitor heart rate, blood oxygen saturation, and body temperature using infrared and photoplethysmography-based sensing technologies. The hardware architecture is based on a flex–rigid design. The device integrates a low-power microcontroller, biometric sensors, and Bluetooth Low Energy (BLE) communication for data transmission to a mobile application, enabling real-time visualisation, historical data logging, support for continuous and manual monitoring modes, and configurable alert threshold notifications. This work represents an initial stage of a broader project, aiming at future system validation and integration into digital health platforms, contributing to more active, safe, and inclusive ageing.