Detection and location of sub-cranial haematomas using mutlifrequency electrical impedance tomograpgy

Head injury is the main cause of death and disability among children and young adults in the UK. When a person experiences a haemorrhage, there is a possibility of developing a haematoma which may or may not have any initial symptoms or external signs. If the patient is not diagnosed and treated in a timely manner, the haematoma can grow and apply pressure to the brain, worsening the neurological condition of the patient, and thus posing a serious threat of paralysis or even death. At present, there is no portable diagnostic/imaging modality for first responders or for continual monitoring of patients for anomalies in the human head. An alternative portable diagnostic system, based on electrical impedance tomography (EIT), is investigated in this thesis.

Multifrequency electrical impedance tomography (MFEIT) is a medical imaging technique that uses surface boundary electrodes to perform measurements and then reconstruct internal conductivity maps of the body under investigation. It is a non-invasive, non-radiating and non-ionising technology that involves the injection of small alternating currents (\(\leq\)10 mA for medical applications) onto the surface of a body using electrocardiogram (ECG) type electrodes and measuring the resulting potentials around the surface. This imaging modality has the potential to be used on conscious, unconscious, or paralysed patients in situations that require either a timely initial diagnosis or continual monitoring.

The research and development of such a system requires bespoke hardware and software for MFEIT, as well as a testing and validation medium other than a live human head in the initial stages. This thesis addresses both of these topics by defining the development of novel hardware that is robust to both frequency and output load changes and a novel geometrically and dielectrically accurate human head phantom model.

The designed MFEIT system has the largest stable combined bandwidth of frequency and current ranges found in the literature, but this achievement of 'flexibility' includes the trade-off of a reduced maximum output impedance and an increased image capture time. This is the first MFEIT system that uses non-paired differential current sources and differential signal to single ended signal converters embedded in to the electrode connector heads. This design decreases the effect of stray noise along the long electrode-to-hardware wires resulting in a system with an average voltage magnitude SNR of 60 dB and a voltage phase SNR of 70 dB.

The testing system necessitates a biofidelic human head phantom model. Existing human head phantom models are very simplistic, treating the brain as a homogeneous tissue and the phantom tissues used only mimic the human head tissues over a narrow frequency range.

The dielectric mimicking materials produced in this research are the most comprehensive and representative set of phantom human head tissues over the frequency range of 100 Hz to 1 MHz. Dielectric mimicking materials are first characterised for the scalp, dura mater, CSF, grey matter, white matter, blood, dentata suture, squamous suture, and dipol\"e bone. These are then used to create a novel human head phantom model with the largest number of distinct tissues that is dielectrically and geometrically accurate over the frequency range of 100 Hz to 1 MHz.

Using the developed MFEIT system with the developed phantom human head model different electrode arrangements and current injection patterns were tested, including a novel semi electrode array. The best results for the detection of anomalies was the opposite method with a 1 kHz frequency difference comparison for an admittivity difference image. For the whole array these images were able to accurately predict the anomaly size and location with few artefacts being produced in the image. The novel semi array results were also able to produce a clear anomaly detection and size estimation, but the anomaly location was affected by the non-uniform sensitivity distribution of the semi array. This produces an anomaly location estimation towards the front of the head rather than at the actual anomaly location. The use of a semi array therefore introduces a trade-off between anomaly location and size accuracy and the practicability of use on trauma patients.