This project aims to investigate a variety of electrodes for ECG (electrocardiogram) measurements and find suitable ones for burn wounded skin in association with the Burn center in Uppsala University Hospital.   To this purpose, the electrical properties (in particular, electrical impedance and equivalent circuit) of electrodes and the influence of the skin on the electrical properties are looked into, and various common artefacts in ECG measurements are investigated, such as wandering baseline (caused by perspiration, respiration, patient movement and poor electrode contact), muscle tremor artefact, 50-60 Hz power line interference and measurement noise.   Simulation of a burn wound was done by putting Ringer’s acetate between two electrodes gel to gel. Six different electrodes made with either a solid or wet gel for the electrolyte were tested, three of which (Ambu Bluesensor L-00-S/25, Ambu Bluesensor R-00-S/25, Milmedtek T-VO01) have wet gel, and three of which (3M 2670-5, Medtronic Arbo, and Ambu Whitesensor WSP30-00-S/50) have solid gel.   The tests showed that the impedance of the electrodes changed as expected and was in almost all cases lowered. An increase in phase shift was also observed with burn wound simulation but could not be proven to relate with increased polarization.   The results showed its wet gel and adhesive developed for sweaty/wet skin, Ambu Bluesensor R-00-S/25 is recommended.   Suggestions for further investigation would be to see if the interference could be solved by impedance balancing the electrodes or to investigate if there is a greater coupling between the wet burn wounds and the main 230V 50Hz network causing higher currents and voltage drops in the body increasing the risk of common mode to differential mode conversion.

Title: Possibilities of monitoring and evaluation of accompanying non-locomotor manifestacions during reflex locomotion according to Vojta Objectives: The aim of study is to determine suitable conditions for measuring and evaluating non-locomotor manifestations. Furthermore, to clarify whether there are changes in the accompanying non-locomotor manifestations during the stimulation of trigger zones from the concept of Vojta's principle. Accompanying non-locomotor manifestations are mainly manifestations of the autonomic nervous system. Measurement of respiratory rate, heart rate and swallowing rate was chosen to evaluate these parameters. Methods: The research was conducted on 7 adult subject for measuring respiratory and heart rate and 12 adult subject for measuring swallowing frequency. These were healthy women aged 18-30. Data were obtained from a CamNtech Actiheard compact ECG sensor and from a video recording. Each proband was first measured for a resting ECG and then measured during activaton of the thoracic trigger zone from the Vojta concept, twice in a row for fifteen minutes. One measurement was performed with the eyes open, the other with the eyes closed, the order was randomized. The obtained data were processed into a video recording and evaluated for each proband separately. It was...

In order to handle the increasing costs of healthcare more of the care and monitoring will take place in the patient’s home. It is therefore desirable to develop smaller and portable systems that can record important biosignals such as the electrical activity of the heart in the form of an ECG. This project is a continuation on a previous project that developed a shield that can be connected to the GPIO pins of a Raspberry Pi, a credit-card sized computer. The shield contains an ADAS1000, a low power and compact device that can record the electrical activity of the heart along with respiration. The aim of this project was to develop an application that can run on the Raspberry Pi in order to display the captured data from the shield on a screen along with storing the data for further processing. The project was successful in the way that the requirements for the software have been fulfilled. / För att hantera den ökande kostnaden för hälso- och sjukvård kommer en större del av övervakning samt vård att ske i patientens hem. Det kommer därför att vara önskvärt att utveckla mindre system som är lättare att hantera än de större traditionella apparaterna för att samla in vanliga biosignaler som exempelvis ett EKG. Detta projekt är en fortsättning på ett tidigare projekt vars syfte var att framställa en ”sköld” som kan kopplas ihop med en Raspberry Pi via dess GPIO pinnar. Det föregående projektet var lyckat och en sköld innehållande en ADAS1000 som kan samla in bl.a. ett EKG samt andningen framställdes. Syftet med detta projekt var att utveckla en applikation som kan köras på en Raspberry Pi och på så sätt visa den data som samlas in från skölden på en skärm. Det skulle även vara möjligt att spara insamlad data för senare användning. Projektet resulterade i en applikation som uppfyllde dessa krav.

In this thesis, many machine learning algorithms were applied to electrocardiogram (ECG), spectral analysis, and Field Programmable Gate Arrays (FPGAs). In ECG, QRS complexes are useful for measuring the heart rate and for the segmentation of ECG signals. QRS complexes were detected using WaveletCNN Autoencoder filters and ConvLSTM detectors. The WaveletCNN Autoencoders filters the ECG signals using the wavelet filters, while the ConvLSTM detects the spatial temporal patterns of the QRS complexes. For the spectral analysis topic, the detection of chemical compounds using spectral analysis is useful for identifying unknown substances. However, spectral analysis algorithms require vast amounts of data. To solve this problem, B-spline neural networks were developed for the generation of infrared and ultraviolet/visible spectras. This allowed for the generation of large training datasets from a few experimental measurements. Graphical Processing Units (GPUs) are good for training and testing neural networks. However, using multiple GPUs together is hard because PCIe bus is not suited for scattering operations and reduce operations. FPGAs are more flexible as they can be arranged in a mesh or toroid or hypercube configuration on the PCB. These configurations provide higher data throughput and results in faster computations. A general neural network framework was written in VHDL for Xilinx FPGAs. It allows for any neural network to be trained or tested on FPGAs. / Graduate

No description available.

The QT interval in an electrocardiogram (ECG) reflects complex processes affecting the repolarization of ventricular myocardium. Increased QT interval variability (QTV) is thought to be caused by ventricular repolarization lability and has been associated with cardiac mortality. Recent publications have shown that template-based methods are more robust than traditional methods for QT interval extraction on a beat-to-beat basis. However, most studies are limited to non-movement ECG recordings, we want to analyze in this study the power of QT interval extraction for mobile non-stationary ECG recordings. The records of 7 test subjects are at least 65 min long and contain about 25 minutes of sport exercise such as running, cycling, sport climbing or acrobatic training. 2DSW was used to extract QT interval and best-fit distance of matched template for signal quality evaluation for each beat. Potential relations between QTV, motion and signal quality are segmentally compared. To determine motion activity we calculated normalized signal magnitude area (SMA). QTV was increased in patients during sport exercise, possibly reflects sympathetic activity in these specific physiological conditions. However, increased QTV could also be caused by low signal quality.

Federated learning is a framework in machine learning that allows for training a model while maintaining data privacy. Moreover, it allows clients with their own data to collaborate in order to build a stronger, shared model. Federated learning is of particular interest to healthcare data, since it is of the utmost importance to respect patient privacy while still building useful diagnostic tools. However, healthcare data can be complicated — data format might differ across providers, leading to unexpected inputs and incompatibility between different providers. For example, electrocardiograms might differ in sampling rate or number of leads used, meaning that a classifier trained at one hospital might be useless to another. We propose using autoencoders to address this problem, transforming important information contained in electrocardiograms to a uniform input, where federated learning can then be used to train a strong classifier for multiple healthcare providers. Furthermore, we propose using statistically-guided hyperparameter tuning to ensure fast convergence of the model.

An electrocardiogram (ECG) is used to monitor the activity of the heart. The human heart beats seventy times on an average per minute. The rate at which a human heart beats can exhibit a periodic variation. This is known as heart rate variability (HRV). Heart rate variability is an important measurement that can predict the survival after a heart attack. Studies have shown that reduced HRV predicts sudden death in patients with Myocardial Infarction (MI). The time interval between each beat is called an RR interval, where the heart rate is given by the reciprocal of the RR interval expressed in beats per minute. For a deeper insight into the dynamics underlying the beat to beat RR variations and for understanding the overall variance in HRV, an accurate method of estimating the RR interval must be obtained. Before an HRV computation can be obtained the quality of the RR interval data obtained must be good and reliable. Most QRS detection algorithms can easily miss a QRS pulse producing unreliable RR interval values. Therefore it is necessary to estimate the RR interval in the presence of missing QRS beats. The approach in this thesis is to apply KALMAN estimation algorithm to the RR interval data calculated from the ECG. The goal is to improve the RR interval values obtained from missed beats of ECG data.

This thesis aims at providing a new approach for detecting R-waves in the ECG signal and generating the corresponding R-wave impulses with the delay between the original R-waves and the R-wave impulses being lesser than 100 ms. The algorithm was implemented in Matlab and tested with good results against 90 different ECG recordings from the MIT-BIH database. The Discrete Wavelet Transform (DWT) forms the heart of the algorithm providing a multi-resolution analysis of the ECG signal. The wavelet transform decomposes the ECG signal into frequency scales where the ECG characteristic waveforms are indicated by zero crossings. The adaptive threshold algorithms discussed in this thesis search for valid zero crossings which characterize the R-waves and also remove the Preventricular Contractions (PVC's). The adaptive threshold algorithms allow the decision thresholds to adjust for signal quality changes and eliminate the need for manual adjustments when changing from patient to patient. The delay between the R-waves in the original ECG signal and the R-wave impulses obtained from the algorithm was found to be less than 100 ms.

Electrocardiogram (ECG) signal processing has been the subject of intense research in the past years, due to its strategic place in the detection of several cardiac pathologies. However, ECG signal is frequently corrupted with different types of noises such as 60Hz power line interference, baseline drift, electrode movement and motion artifact, etc. In this thesis, a hybrid two-stage model based on the combination of wavelet decomposition and artificial neural network is proposed for ECG noise reduction based on excellent localization features: wavelet transform and the adaptive learning ability of neural network. Results from the simulations validate the effectiveness of this proposed method. Simulation results on actual ECG signals from MIT-BIH arrhythmia database [30] show this approach yields improvement over the un-filtered signal in terms of signal-to-noise ratio (SNR).