Anomaly Detection from BCG Signal With Artificial Intelligence
- Typ der Arbeit: Forschungsprojekt
- Status der Arbeit: abgeschlossen
- Projekte: PotatoNet
- Betreuer: Ulf Kulau
- Ende der Arbeit: 27. Jan 2023
From recent work it can be summarized that deep learning has been performing reliably and accurately in detecting anomalies from ECG data. According to the current state of the art, automatic anomaly detection from BCG data is still lacking. Thus, in this work, we will use deep learning to detect anomalies from BCG data.
Research will be done more specifically on the BCG data and anomaly detection with deep learning. The workflow of the implementation will be finalized and required tools will be determined.
BCG data will be collected and Exploratory Data Analysis will be performed. The collected data will be processed and used as training data. As we do not have BCG data from heart disease patients, we will generate test data by separating a part of the training data and injecting erroneous data into it.
As deep learning model development is an iterative process combined with develop- ment, training, testing, and evaluation, we will spend most of the time in this work package. We will continuously improve the model by tuning model architecture, layers, and parameters and determine the best-performed model.
- Data analysis and pre-processing: Analysis of the dataset. Find out the model’s relevant feature. Generate training and test dataset
- Model development: We will develop deep learning-based autoencoder models. We will iteratively train the model with our dataset for different layers, activation functions, and parameters.
- Evaluation of the model: Evaluate the models with different performance parameters and determine the best-performed model.
- Ballistocardiography and seismocardiography: A review of recent advances
IEEE journal of biomedical and health informatics19.4IEEE2014.
- Estimation of central aortic forces in the ballistocardiogram under rest and exercise conditions
2009 Annual International Conference of the IEEE Engineering in Medicine and Biology Society2009.
- Cost-Sensitive Learning for Anomaly Detection in Imbalanced ECG Data Using Convolutional Neural Networks
Sensors22.11Multidisciplinary Digital Publishing Institute2022.