Engineering

Publication Search Results

Sort
Results per page
Now showing 1 - 2 of 2
  • Optical-Electrode: The Next Generation Brain-Machine Interface
    (2022) Wei, Yuan
    Thesis
    Brain machine interfaces, or brain computer interfaces, are attracting ever increasing research interests for their promising application prospects. A number of methods and devices were proposed on this topic, but all have inherent limits particularly concerning spatial density and signal resolution. An optical-electrode is hereby proposed to overcome these limitations by transducing the electrical signal into an optical signal using liquid crystal cells. In addition, photovoltaic stimulation capabilities were added to form an integrated bidirectional interface. A recording subsystem and a stimulating subsystem were proposed for driving the sensing and stimulation parts respectively, and their benchtop characterisations were carried out. Noise performances in the recording subsystem were analysed and optimised. To provide initial validation, animal studies were conducted on rabbit sciatic nerves (in vivo and ex vivo) and on cardiac tissues (ex vivo). The recorded signals and stimulated responses were compared with those made by commonly used traditional electrical systems under the same experimental conditions. Compound action potentials, although showing differences on delays and morphology over traditional methods, were successfully recorded and evoked. The charge balance ability was also demonstrated in the experiments. Finally, a 'zero mode' photodetector is introduced, which is specifically suitable for the recording subsystem and can potentially improve the noise performance. The works in this thesis will contribute to the next iteration of the technology, i.e. help the creation of high density arrays in the form of integrated chips.
  • Deep learning methods for the estimation of preload and physiological control of heart assist devices
    (2021) Fetanat Fard Haghighi, Masoud
    Thesis
    Heart failure (HF) is one of the most prevalent life-threatening cardiovascular diseases, affecting more than 23 million worldwide. Although heart transplantation is the gold standard treatment for end-stage HF patients, the number of donor hearts is significantly less than the demand. Mechanical circulatory support for a patient with a failing left ventricle can be achieved by implanting a left ventricular assist device (LVAD) by pumping blood from the left ventricle to the aorta. Currently, clinicians set the LVAD speed at a fixed value, which can lead to different hazardous events. A physiological control system (PCS), which automatically adjusts pump speed can mitigate the hazardous events and improve a patient’s mobility, lifespan and quality of life. However, there are two main reasons that the current PCSs are not used commercially. Firstly, previously developed PCSs have been evaluated in specific conditions for only single-patient scenarios. Secondly, previously developed PCSs require implanted pressure or flow sensors. Therefore, the aim of this thesis was to design novel methods for estimation of preload and sensorless PCS for LVADs than can accommodate interpatient and intrapatient variations (IAIV), by way of three objectives. The first objective was to design a PCS for an implantable heart pump that accommodates IAIV. A novel model free adaptive control (MFAC) system was developed that maintained the preload in the normal range of 3 to 15 mmHg for different patient conditions. The second objective was to design a sensorless PCS for LVADs across different patient conditions, by combining a preload estimator using a deep learning method and the MFAC. The third objective was to design an improved non-invasive preload estimator, based on deep learning methods using LVAD flow waveforms recorded clinically. The proposed preload estimator was extremely accurate with a correlation coefficient of 0.97 and root mean squared error of 0.84 mmHg. The proposed sensorless PSC works similarly to the preload-based PCS using measured preload to prevent suction and congestion. This study shows that the LVADs can respond appropriately to changing patient states and physiological demands without the need for additional implanted sensors.