Signal Processing-Driven AI for Healthcare
- 1 Edición - 1 de octubre de 2026
- Última edición
- Autores: Ildar Rakhmatulin, Ganesh R. Naik
- Idioma: Inglés
Signal Processing-Driven AI for Healthcare examines how AI techniques can be applied across four major biosignals—EEG, EMG, EOG, and ECG—to derive clinically meaningful insight… Leer más
Descripción
Descripción
Signal Processing-Driven AI for Healthcare examines how AI techniques can be applied across four major biosignals—EEG, EMG, EOG, and ECG—to derive clinically meaningful insights. As biomedical data becomes increasingly multimodal, there is a rising need for integrated methodologies that unite these signals within robust, explainable AI pipelines suitable for healthcare environments. This book provides a unified framework that spans data acquisition, preprocessing, feature extraction, modeling, evaluation, and deployment, with an emphasis on reproducibility, practical Python-based implementations, and real-world translation to clinical workflows.
Puntos claves
Puntos claves
- Integrates AI-driven signal processing across EEG, EMG, EOG, and ECG
- Presents end-to-end workflows from data acquisition to deployment
- Demonstrates multimodal fusion and clinical decision support
- Emphasizes interpretability, validation, and regulatory considerations
- Features supplementary website to host the dataset
- Offers ready-to-use Python notebooks and scripts optimized for Google Collaboration
De interès para
De interès para
Researchers, graduate students, and professionals in biomedical engineering and computer science focusing on signal processing and analysis
Índice
Índice
Part I — Foundations
1. Introduction: AI, biosignals, and healthcare workflows
1.1 What are biosignals? (EEG, EMG, EOG, ECG — definitions & clinical roles)
1.2 Why AI for biosignals? Opportunities & limitations
1.3 Overview of the signal→feature→model pipeline
1.4 Data, privacy, and regulatory context in healthcare
2. Sensors, acquisition, and data characteristics
2.1 Electrodes and sensor types (scalp, surface, invasive)
2.2 Sampling, quantization, and anti-aliasing
2.3 Typical noise sources & artifacts (powerline, motion, ocular, muscle)
2.4 Multi-channel setups and montage conventions (EEG 10–20, ECG leads, EMG placements)
2.5 Public datasets and benchmark repositories (PhysioNet, Sleep-EDF, TUH EEG, etc.)
Part II — Signal Processing for Biomedical Time Series
3. Preprocessing & cleaning
3.1 Filtering basics: high/low/bandpass, notch filters
3.2 Baseline wander and drift correction
3.3 Artifact detection & removal: ICA, SSP, regression approaches
3.4 Channel interpolation & re-referencing
3.5 Quality metrics and pipeline reproducibility
4. Time-frequency and feature transforms
4.1 Fourier analysis and spectral features
4.2 Short-time Fourier transform (spectrograms)
4.3 Wavelets and multi-resolution analysis
4.4 Empirical mode decomposition & Hilbert spectrum
4.5 Feature normalization and selection techniques
5. Spatial and multichannel processing
5.1 Spatial filters: common average, Laplacian
5.2 Source localization basics (inverse problems, dipole models)
5.3 Spatial-spectral feature extraction (CSP for BCI)
5.4 Dimensionality reduction: PCA, ICA, manifold learning
Part III — Machine Learning & Deep Learning Methods
6. Classical ML for biosignals
6.1 Feature engineering strategies (time, frequency, non-linear features)
6.2 Common classifiers: SVMs, Random Forests, k-NN, logistic regression
6.3 Model evaluation: cross-validation, time-series splits, metrics (sensitivity, specificity, AUC)
6.4 Imbalanced data, augmentation, and synthetic data (SMOTE, GANs)
7. Deep learning approaches
7.1 CNNs for spectrograms and spatial patterns
7.2 RNNs, LSTMs, and temporal modeling
7.3 Transformers for sequence modeling: adaptations for biosignals
7.4 Hybrid architectures and multi-branch networks
7.5 Transfer learning, pretraining, and fine-tuning
8. Explainability, uncertainty, and interpretability
8.1 Saliency maps, Grad-CAM, and attention visualizations
8.2 Model calibration and uncertainty quantification
8.3 Causal reasoning and clinically meaningful explanations
8.4 Human-in-the-loop and clinician trust
Part IV — Modality-Focused Chapters (EEG, EMG, EOG, ECG)
9. EEG: brain signals, pipelines, and applications
9.1 EEG physiology and rhythms (delta, theta, alpha, beta, gamma)
9.2 Preprocessing specific to EEG (ocular and muscle artifact handling)
9.3 Applications: seizure detection, sleep staging, BCI, cognitive state monitoring
9.4 Case study: automated sleep staging with deep CNNs (dataset, code, results)
9.5 Practical tips: montages, channel selection, subject variability
10. ECG: cardiac signal analytics and arrhythmia detection
10.1 ECG morphology and clinical intervals (P, QRS, T, PR, QT)
10.2 Beat detection, HRV features, and noise robustness
10.3 Applications: arrhythmia classification, ischemia detection, wearables
10.4 Case study: real-time arrhythmia detection on wearable device (edge considerations)
10.5 Regulatory & clinical-trial considerations for cardiac AI
11. EMG: muscle activation, prosthetics, and fatigue monitoring
11.1 Physiology of EMG; surface vs. intramuscular recordings
11.2 Feature sets for EMG classification (time-domain, spectral, envelope)
11.3 Applications: prosthetic control, motor disorder diagnosis, fatigue detection
11.4 Case study: EMG-driven prosthetic control with CNN+LSTM
11.5 Practicalities: electrode placement, cross-talk, normalization
12. EOG: eye movement, drowsiness, and human factors
12.1 EOG basics: saccades, blinks, slow eye movements
12.2 Removing EOG artifacts from EEG and using EOG as a signal itself
12.3 Applications: drowsiness detection, gaze estimation, assistive tech
12.4 Case study: driver drowsiness detection from EOG + accelerometer fusion
Part V — Multimodal Fusion, Deployment & Systems
13. Multimodal learning and sensor fusion
13.1 Why fuse? Complementary signals & improved robustness
13.2 Early fusion vs. late fusion strategies
13.3 Synchronization, alignment, and missing data handling
13.4 Example: EEG+ECG for seizure vs. syncope differentiation
14. Real-time systems, edge AI, and hardware considerations
14.1 Latency, throughput, and power tradeoffs
14.2 Model compression: pruning, quantization, distillation
14.3 Embedded inference (microcontrollers, mobile GPUs, NPUs)
14.4 Streaming architectures and pipelining for continuous monitoring
15. Data engineering, annotation, and labelling strategies
15.1 Creating reliable labels: clinical annotation workflows
15.2 Weak labels, semi-supervised learning, and self-supervised pretraining
15.3 Data versioning, provenance, and reproducible pipelines
1. Introduction: AI, biosignals, and healthcare workflows
1.1 What are biosignals? (EEG, EMG, EOG, ECG — definitions & clinical roles)
1.2 Why AI for biosignals? Opportunities & limitations
1.3 Overview of the signal→feature→model pipeline
1.4 Data, privacy, and regulatory context in healthcare
2. Sensors, acquisition, and data characteristics
2.1 Electrodes and sensor types (scalp, surface, invasive)
2.2 Sampling, quantization, and anti-aliasing
2.3 Typical noise sources & artifacts (powerline, motion, ocular, muscle)
2.4 Multi-channel setups and montage conventions (EEG 10–20, ECG leads, EMG placements)
2.5 Public datasets and benchmark repositories (PhysioNet, Sleep-EDF, TUH EEG, etc.)
Part II — Signal Processing for Biomedical Time Series
3. Preprocessing & cleaning
3.1 Filtering basics: high/low/bandpass, notch filters
3.2 Baseline wander and drift correction
3.3 Artifact detection & removal: ICA, SSP, regression approaches
3.4 Channel interpolation & re-referencing
3.5 Quality metrics and pipeline reproducibility
4. Time-frequency and feature transforms
4.1 Fourier analysis and spectral features
4.2 Short-time Fourier transform (spectrograms)
4.3 Wavelets and multi-resolution analysis
4.4 Empirical mode decomposition & Hilbert spectrum
4.5 Feature normalization and selection techniques
5. Spatial and multichannel processing
5.1 Spatial filters: common average, Laplacian
5.2 Source localization basics (inverse problems, dipole models)
5.3 Spatial-spectral feature extraction (CSP for BCI)
5.4 Dimensionality reduction: PCA, ICA, manifold learning
Part III — Machine Learning & Deep Learning Methods
6. Classical ML for biosignals
6.1 Feature engineering strategies (time, frequency, non-linear features)
6.2 Common classifiers: SVMs, Random Forests, k-NN, logistic regression
6.3 Model evaluation: cross-validation, time-series splits, metrics (sensitivity, specificity, AUC)
6.4 Imbalanced data, augmentation, and synthetic data (SMOTE, GANs)
7. Deep learning approaches
7.1 CNNs for spectrograms and spatial patterns
7.2 RNNs, LSTMs, and temporal modeling
7.3 Transformers for sequence modeling: adaptations for biosignals
7.4 Hybrid architectures and multi-branch networks
7.5 Transfer learning, pretraining, and fine-tuning
8. Explainability, uncertainty, and interpretability
8.1 Saliency maps, Grad-CAM, and attention visualizations
8.2 Model calibration and uncertainty quantification
8.3 Causal reasoning and clinically meaningful explanations
8.4 Human-in-the-loop and clinician trust
Part IV — Modality-Focused Chapters (EEG, EMG, EOG, ECG)
9. EEG: brain signals, pipelines, and applications
9.1 EEG physiology and rhythms (delta, theta, alpha, beta, gamma)
9.2 Preprocessing specific to EEG (ocular and muscle artifact handling)
9.3 Applications: seizure detection, sleep staging, BCI, cognitive state monitoring
9.4 Case study: automated sleep staging with deep CNNs (dataset, code, results)
9.5 Practical tips: montages, channel selection, subject variability
10. ECG: cardiac signal analytics and arrhythmia detection
10.1 ECG morphology and clinical intervals (P, QRS, T, PR, QT)
10.2 Beat detection, HRV features, and noise robustness
10.3 Applications: arrhythmia classification, ischemia detection, wearables
10.4 Case study: real-time arrhythmia detection on wearable device (edge considerations)
10.5 Regulatory & clinical-trial considerations for cardiac AI
11. EMG: muscle activation, prosthetics, and fatigue monitoring
11.1 Physiology of EMG; surface vs. intramuscular recordings
11.2 Feature sets for EMG classification (time-domain, spectral, envelope)
11.3 Applications: prosthetic control, motor disorder diagnosis, fatigue detection
11.4 Case study: EMG-driven prosthetic control with CNN+LSTM
11.5 Practicalities: electrode placement, cross-talk, normalization
12. EOG: eye movement, drowsiness, and human factors
12.1 EOG basics: saccades, blinks, slow eye movements
12.2 Removing EOG artifacts from EEG and using EOG as a signal itself
12.3 Applications: drowsiness detection, gaze estimation, assistive tech
12.4 Case study: driver drowsiness detection from EOG + accelerometer fusion
Part V — Multimodal Fusion, Deployment & Systems
13. Multimodal learning and sensor fusion
13.1 Why fuse? Complementary signals & improved robustness
13.2 Early fusion vs. late fusion strategies
13.3 Synchronization, alignment, and missing data handling
13.4 Example: EEG+ECG for seizure vs. syncope differentiation
14. Real-time systems, edge AI, and hardware considerations
14.1 Latency, throughput, and power tradeoffs
14.2 Model compression: pruning, quantization, distillation
14.3 Embedded inference (microcontrollers, mobile GPUs, NPUs)
14.4 Streaming architectures and pipelining for continuous monitoring
15. Data engineering, annotation, and labelling strategies
15.1 Creating reliable labels: clinical annotation workflows
15.2 Weak labels, semi-supervised learning, and self-supervised pretraining
15.3 Data versioning, provenance, and reproducible pipelines
Detalles del producto
Detalles del producto
- Edición: 1
- Última edición
- Publicado: 1 de octubre de 2026
- Idioma: Inglés
Sobre los autores
Sobre los autores
IR
Ildar Rakhmatulin
Dr. Ildar Rakhmatulin is a scientist and the creator of several popular open-source brain-computer interface (BCI) projects. He is the founder of PiEEG, a low-cost BCI solution. His experience includes working as a BCI developer at Imperial College London, a machine learning researcher at Heriot-Watt University, and a researcher in the neurotechnology group at the University of Edinburgh, UK. Additionally, he is the author of neuroscience courses on Udemy.
Afiliaciones y experiencia
Independent Engineer and Founder, PiEEG Solutions, Edinburgh, UKGN
Ganesh R. Naik
Dr. Ganesh R. Naik is a leading researcher in biomedical engineering and signal processing, ranked among the top 2% of scientists globally (Stanford University). He earned his PhD in Electronics Engineering from RMIT University, Melbourne (2009), and is currently a Senior Academic and Researcher in Computer Science and IT at Torrens University Australia. Dr. Naik has edited 16 books and published over 150 peer-reviewed papers. He serves as Associate Editor for IEEE Access, Frontiers in Neurorobotics, and two Springer journals. His career is distinguished by fellowships from Baden–Württemberg (Germany), ISSI (Australia), the BridgeTech Program, and the Royal Academy of Engineering (UK). Previously, he held research roles at Flinders University, Western Sydney University, and UTS, contributing to major projects in sleep technology, wearable sensors, and AI-driven biomedical signal processing.
Afiliaciones y experiencia
Torrens University, Adelaide, Australia