PREdictive analytics for machine DIagnositcs and healthCare applicaTions
The IDLab-PreDiCT team focuses on the study and design of innovative machine learning and hybrid AI solutions with applications in predictive maintenance and predictive healthcare.
The IDLab-PreDiCT team focuses on following lines of research:
- Design of (context-aware) machine learning and hybrid AI on
- Time series data
- Wearable and/or smartphone sensor data
- Example use cases: human activity recognition, stress monitoring, migraine management, cognitive decline in MS, early depression relapse detection, homecare monitoring, e-care, nurse call assessment & assignment, anxiety detection
- Speech
- Example use cases: stress/pain detection in free or semi-guided speech
- Neuromodulation
- Example use case: chronic pain management, movement classification, epilepsy detection, TMS
- Industrial monitoring sensors (e.g. accelerometer, pressure, power)
- Example use cases: anomaly detection, fault detection, remaining useful life prediction, wind power prediction, leak detection and localisation
- Building monitoring sensors (e.g. CO2, VOC, humidity)
- Example use cases: comfort assessment, building management, optimization HVAC, energy consumption, renovation suggestion
- Tool design for optimized processing
- Series Distance Matrix
- Plotly-resampler
- Tsflex
- Structured data
- EHR
- Example use cases: AF, COPD progression or regression detection, infection management, personalized therapy for psoriasis
- Process data / decision trees / protocols / guidelines
- Example use cases: chemical engineering, fault and anomaly detection, outcome prediction
- Semantic data
- Ontology design workshops
- Example use cases: Fault and Anomaly Detection for industry 4.0, processes for chemical engineering, tacit knowledge mapping in healthcare
- Semantic rule mining
- Example use cases: Trigger detection for chronic diseases, process optimization, correlation detection for healthcare
- Knowledge (Graph) embedding
- Example use cases: Mortality prediction, network monitoring, cross-context learning, household activity recognition, chemical engineering, disease follow-up
- Graph-based machine learning, e.g. GCN
- Example use cases: Mortality prediction, link prediction, node classification
- Optimal architectures for fusing ML & Semantics (Hybrid AI architectures)
- Image data
- Human and animal health diagnosis & prognostics
- CT / x-ray / RGB / scan slices
- Example use cases: hip dysplasia / TAVR / tuberculosis / wound progression / burn wound management / parasite detection
- Machine monitoring
- RGB / IR
- Example use cases: weld fault detection
- Multi-modal and multi-sensor data
- Combining structured data and time series, or structured data and images, or all three
- Example use cases: TAVR, Cataract
- Decision support and interventions
- EMA and behavioral monitoring app
- Social chatbots
- Example use cases: smoking cessation
- Intervention management
- Example use cases: smoking cessation, nurse call assignment, digital health marker design & follow-up, personalized anxiety therapy
- Dynamic dashboards
- User-driven & event-driven visualization
- Visualization recommendation
- Feedback gathering and conflict resolution to optimize AI/ML
- Design of explainable and trustworthy AI
- Explainability methods
- Feature importance methods
- Uncertainty quantification
- Correlation mining
- Causal AI
- Privacy enhancing solutions: ML on SOLID, federated learning
IDLab-PreDiCT performs basic, applied and contract research on the above topics. Our expertise has been successfully used by academic and business partners worldwide. We work closely with technology leaders on real world challenges.
More detailed information on our activities can be found on our IDLab-PreDiCT team website.
