Biomedical Engineering Systems and Technologies: 15th International Joint Conference, BIOSTEC 2022, Virtual Event, February 9–11, 2022, Revised ... in Computer and Information Science)

This book constitutes the refereed post-proceedings of the 15th International Conference on Biomedical Engineering Systems and Technologies, BIOSTEC 2022, held as a Virtual Event, during February 9–11, 2022. The 21 full papers included in this book were carefully reviewed and selected from 262 submissions. The papers selected to be included in this book contribute to the understanding of relevant trends of current research on Biomedical Engineering Systems and Technologies, including: Pattern Recognition and Machine Learning, Application of Health Informatics in Clinical Cases, Evaluation and Use of Healthcare IT, Medical Signal Acquisition, Analysis and Processing, Data Mining and Data Analysis, Decision Support Systems, e-Health, e-Health Applications, Mobile Technologies for Healthcare Applications and Medical Devices design. Preface Organization Contents First Version of a Support System for the Medical Diagnosis of Pathologies in the Larynx 1 Introduction 2 Materials 2.1 Database 2.2 Voice Pathologies 3 Methods 3.1 Feature Extraction 3.2 Classification Procedure 4 Results and Discussion 5 Conclusions References Analysis of Extracellular Vesicle Data on Fluorescence and Atomic Force Microscopy Images 1 Introduction 1.1 Goal 2 State of the Art 3 Methods 3.1 Determination of Green Fluorescent Proteins in Extracellular Vesicles 3.2 Quality Assessment of Extracellular Vesicle Populations 4 Results 4.1 Determination of Green Fluorescent Protein in Extracellular Vesicle 4.2 Quality Assessment of Extracellular Vesicle Populations 5 Conclusion References Automated Segmentation of Patterned Cells in Micropatterning Microscopy Images 1 Background 2 State of the Art 3 Methodology 3.1 Data and Material 3.2 Gridding 3.3 Automated Cell Segmentation 4 Results 4.1 Segmentation Results 4.2 Model Training 4.3 Prediction Results 4.4 Post-processing 5 Conclusion and Outlook References Automated Data Adaptation for the Segmentation of Blood Vessels 1 Introduction 1.1 Problem Description 1.2 State of the Art 2 Methods 2.1 Heuristic Image Manipulation 2.2 Model Based Pre-processing 3 Results 3.1 Evolutionary Preprocessing 3.2 Transfer Learning and Retraining 3.3 Recoloring Neural Network 4 Conclusion References Security Analysis of the Internet of Medical Things (IoMT): Case Study of the Pacemaker Ecosystem 1 Introduction 2 Background 2.1 The Pacemaker Ecosystem 2.2 Related Research 2.3 Threat Model 3 Methodology 3.1 Targets 3.2 Black Box Testing 3.3 Hardware Testing Methodology and Setup 3.4 Network Testing Setup 3.5 Modem Fuzzing Setup 3.6 Ethical Considerations 4 Security Analysis 4.1 Hardware 4.2 Firmware 4.3 Communication 4.4 Infrastructure 4.5 General Considerations and Attack Scenarios 5 Discussion 5.1 Results 5.2 Trade-Offs in the Medical Industry 5.3 Device Management and Credentials Invalidation 5.4 Mitigation and Defense 5.5 New Regulations 6 Conclusion References Parallel Lossy Compression for Large FASTQ Files 1 Introduction 2 Background 3 Noise Reduction of DNA Sequences and Smoothing Quality Scores 4 Parallel Strategy and Paired-End Mode 5 Experiments 6 Effects on Variant Calling 7 Conclusions and Discussion References Comparing Different Dictionary-Based Classifiers for the Classification of Volatile Compounds Measured with an E-nose 1 Introduction 2 Related Work 2.1 E-nose for Disease Diagnosis 2.2 Signal Classification 3 Acquisition Setup and Dataset 4 Methods 4.1 Pre-processing 4.2 SSTS for Time Series Text Representation 4.3 Pattern Search and Sentence Generation 4.4 Text Vectorization Methods 5 Results and Discussion 5.1 Overall Results 5.2 Dependence of the Formulation to Classify VOCs 5.3 Performances of Text-Mining Methods 6 Conclusion and Future Work References High-Level Features for Human Activity Recognition and Modeling 1 Background and Related Works 2 High-Level Features 2.1 Concept 2.2 Proposed High-Level Features 2.3 HLF Assignments 3 Feature Extraction 4 Activity Classification 5 Error Attribution 6 Few-Shot Learning 6.1 Imbalanced Learning 6.2 Few-Shot Learning 7 Dataset Combination 8 Future Work 9 Conclusion References Data Augmentation Based on Virtual Wrist Devices for Fall Detection 1 Introduction 2 Related Work 3 Approach 3.1 Feature Computation 3.2 Classifiers 4 Datasets 4.1 Wrist Device (WD) - Arduino-Based 4.2 Virtual Device from Characters in Unity (VDC) 4.3 Virtual Device from Simulations in Unity (VDS) 5 Experiments 5.1 Baseline Model and Results from ch9biosignals22 5.2 Data Augmentation Tests - Testing Set from WD Data 6 Conclusions References Propagation of Response Signals Registered in EEG Under Photostimulation 1 Introduction 2 Dispersion Phenomenon in the Propagation of Disturbances in Biological Media 3 EEG Bursts Propagating Under Photostimulation 4 Discussion and Conclusion References Mobile Tele-Dermatology Use Among University Students: A Pilot Study at Saint Joseph University (USJ) 1 Introduction 2 Approach 2.1 Summary of Our Statistical Pilot Study 2.2 Conclusions from Our Statistical Analysis 2.3 Sample Overview 3 Descriptive Analysis Results 3.1 Social Determinants - Age and Gender 3.2 Social Determinants - Age and Marital Status 3.3 Education Specialty 3.4 Result Demonstrability 3.5 Perceived Risk 3.6 Subjective Norms 3.7 Medical Factors 4 Summary of Findings 5 Conclusion Appendix Survey Questions and Possible Answers Latent Variables for Our Statistical Model from Aroutine et al. (2022) Hypotheses - Summarized from Aroutine et al. (2022) References Predictive Alarm Prevention by Forecasting Threshold Alarms at the Intensive Care Unit 1 Introduction 2 Data Preparation 2.1 Data Slicing 2.2 Data Cleaning 2.3 Extracting Alarm Events 2.4 Resampling Vital Parameters 2.5 Chunking to Avoid Data Gaps 3 Alarm Forecasting 4 Results 5 Discussion References ST-Segment Anomalies Detection from Compressed Sensing Based ECG Data by Means of Machine Learning 1 Introduction 2 Background and Related Work 2.1 ST-Related Conditions 2.2 Automatic Detection of ST-Related Conditions 2.3 The Compressed Sensing Algorithm 3 Automatically Detecting ST-Related Conditions from Compressed ECG: Workflow of the Approach 3.1 Introducing RASTC: Uncompressed vs Compressed Domain 3.2 ECG Signal Compression 3.3 Features Evaluated on the Compressed ECG Information 3.4 Beat Classification 4 The Study 4.1 Study Design 4.2 Study Results 5 Conclusion and Future Works References A Proof-of-Concept Implementation Based on the Framework of AI-Enabled Proactive mHealth: Health Promotion with Motivation 1 Introduction 2 Related Work 3 The Framework of AI-Enabled Proactive mHealth with Modules 3.1 Module-1 -Rules- Decision-Making and Context-Awareness 3.2 Module-2 - Tools-AI Capabilities with Automated Decision-Making and Predictive Analytics 3.3 Module-3 Design with P5 Approach to Mhealth 3.4 Module-4 the Architecture with Just-in-Time Adaptive Interventions 3.5 Module-5- Implementation with Components, Parameters, Factors, and Features: Sources, Wearables, Data Points and APIS 4 Health Interventions with Categorization and Properties 5 Proof-of-Concept Implementation 6 Discussion 7 Limitation 8 Conclusions and Future Work References Improved Blood Vessels Segmentation of Infant Retinal Image 1 Introduction 2 Related Work 3 Method and System Design 3.1 Datasets and Material 3.2 Image Pre-processing 3.3 Blood Vessels Segmentation 3.4 Vessels Segmentation and DCNN Architecture 4 Experimental Results 4.1 Evaluation Metrics 4.2 Preprocessing of Fundus Images 4.3 DL-Model Training and Testing 4.4 Blood Vessels Segmentation 4.5 Various Combinations for Vessel Extraction Models 4.6 Applications of Blood Vessel Segmentation 5 Conclusion References On the Impact of the Vocabulary for Domain-Adaptive Pretraining of Clinical Language Models 1 Introduction 2 Related Work 3 Data 3.1 Pretraining Data 3.2 Fine-Tuning Data 4 Methods 4.1 Baseline – Inheriting the General-Domain Vocabulary 4.2 Vocabulary Development 4.3 Domain-Adaptive Pretraining 4.4 Downstream Task Fine-Tuning 5 Results 6 Discussion 7 Conclusions References A Systematic Literature Review of Extended Reality Exercise Games for the Elderly 1 Introduction 2 Related Work 3 Systematic Literature Review 4 Results 4.1 Publication Information and Keywords 4.2 Immersive Technologies and Game Concepts 4.3 Teamwork and Social Games 4.4 Evaluation 4.5 Opportunities and Challenges 4.6 Adapting Design 5 Discussion 6 Conclusions and Future Work References Simulating the Vital Signs of a Digital Patient Undergoing Surgery, for the Purpose of Training Anaesthetists 1 Introduction 2 Related Work 3 Case-Based Reasoning Framework 3.1 Contextualized Multidimensional Pattern 3.2 Outline of Algorithm 3.3 Identifying the k Nearest Neighbours of the Digital Patient 3.4 Predicting the Short-Term Evolution of the Digital Patient 4 Evaluation of the SVP-OR Simulator 4.1 Data Set 4.2 Selection of the Univariate Dissimilarity Measure 4.3 Implementation and Parameterization 4.4 Experimental Set Up to Assess the Realism of the SVP-OR Simulations 4.5 Results and Discussion 5 Conclusion and Future Work References A Multi-Modal Dataset (MMSD) for Acute Stress Bio-Markers 1 Introduction 2 Physiology of Stress 3 Experimental Setup 3.1 Hardware 3.2 Software and Stressors 4 Eligibility Criteria and Participants 5 Experimental Design 6 Protocol Validation and Labeling 7 Collected Data 7.1 Signal Processing and Feature Extraction 8 Statistical Analysis 9 Case Study Strengths and Limitations References On the Use of WebAssembly for Rendering and Segmenting Medical Images 1 Introduction 2 Stone of Orthanc 2.1 Motivations 2.2 Loaders and Oracle 2.3 Viewports and Layers 2.4 Sample Desktop Viewer 3 Stone Web Viewer 4 Client-Side Segmentation Using Deep Learning 4.1 U-Net Architecture 4.2 Dataset for Lung Segmentation and Data Augmentation 4.3 Training 4.4 U-Net Models in the Stone Web Viewer 5 Other Applications 5.1 Rendering Oncology Images 5.2 Real-Time Volume Reslicing 5.3 Digitally Reconstructed Radiographs 6 Conclusions References Referable Diabetic Retinopathy Detection Using Deep Feature Extraction and Random Forest 1 Introduction 2 Background 2.1 Deep Learning Techniques for Feature Extraction 2.2 Classification Techniques 3 Related Works 4 Data Preparation 4.1 Data Acquisition 4.2 Data Preprocessing and Augmentation 5 Experimental Process 6 Results and Discussion 6.1 (RQ1): Do Hybrid RF Models Perform Better than Their Trees? 6.2 (RQ2): What Is the Best Number of Trees for Each Hybrid RF Model over Each Feature Extractor? 6.3 (RQ3): Do Hybrid RF Models Perform Better than the DT Classifiers? 6.4 (RQ4): What Is the Best Hybrid RF Model for Each Dataset and over the Three Datasets? 7 Threats of Validity 8 Conclusion and Future Work References Author Index