Nanotechnology has the potential to revolutionise every facet of human life. One particularly exciting branch of nanotechnology involves the construction of nanodevices using protein cages. Protein cages are spherically shaped structures with large internal cavities. The research described in this thesis was conducted with the aim of rationalising the design and fabrication of protein cage-based nanodevices. Protein-based nanodevices are typically constructed by re-engineering naturally occurring protein chassis (e.g. ferritin). To rationalise the process of chassis selection, an online registry of protein cages, rings and tubes was designed and populated by computationally mining the Protein Data Bank. The resulting registry was made publically available to the research community through the website – www.nanodevice.build. The functionality of protein cage-based nanodevices can be augmented by packaging inorganic nanoparticles inside their internal cavities. The methods currently used to achieve this typically involve exposure to harsh conditions, which can cause irreversible damage to the protein cage. To address this, a strategy for efficiently packaging inorganic nanoparticles into protein cages under mild conditions was formulated and tested. These experiments were conducted using gold nanoparticles and a number of different protein cages (e.g. Bfr, FtnH and FtnL). Cholangiocarcinoma (CCA) is a deadly liver cancer for which current treatment options are limited. Therefore a CCA-targeting protein cage-based nanodevice was designed, constructed and experimentally evaluated. CCA-targeting was achieved in the context of the CCA cell line TFK-1 using an anti-mesothelin antibody as a targeting agent. Collectively, these three outputs provide a rational framework for selecting a protein cage chassis, loading it with a pre-fabricated inorganic nanoparticle and targeting the resulting device to a particular cell-type. It is hoped that by leveraging these three tools, synthetic biologists will be able to engineer a new generation of nanodevices.

Predicting suicide risk for mental health patients is a challenging task performed by practitioners on a daily basis. Failure to perform proper evaluation of this risk could have a direct effect on the patient's quality of life and possibly even lead to fatal outcomes. Risk predictions are based on data that are difficult to analyse because they involve a heterogeneous set of patients’ records from a high-dimensional set of potential variables. Patient heterogeneity forces the need for various types and numbers of questions to be asked regarding the individual profile and perceived level of risk. It also results in records having different combinations of present variables and a large percentage of missing ones. Another problem is that the data collected consist of risk judgements given by several thousand assessors for a large number of patients. The problem is how to use the associations between patient profiles and clinical judgements to generate a model that reflects the agreement across all practitioners. In this thesis, a novel dynamic feature selection algorithm is proposed which can predict the risk level based only on the most influential answers provided by the patient. The feature selection optimises the vector for predictions by selecting variables that maximise correlation with the assessors’ risk judgement and minimise mutual information within the ones already selected. The final vector is then classified using a linear regression equation learned for all patients with a matching set of variables. The overall approach has been named the Dynamic Feature Selection and Prediction algorithm, DFSP. The results show that the DFSP is at least as accurate or more accurate than alternative gold-standard approaches such as random forest classification trees. The comparison was based on accuracy and error measures applied to each risk level separately ensuring no preference to one risk over the other.

Aligned fibrillar scaffolds (AFSs) have been widely studied for their application in regenerative medicine, providing possible transplantable tissue replacements for nerve, spinal cord, tendon, ligament, muscle, etc. However, researches in AFSs are technically challenging mainly due to the complex fabrication and characterization processes, especially when the AFSs are made to be fully three-dimensional (3D). As the structure is linked to the quality and function of the engineered tissue product, there is an urgent need for novel techniques to characterize AFSs non-invasively and non-destructively and to link their characteristics to their functions and outcome. In this thesis AFS fabrication and characterization were explored. By combining second harmonic generation (SHG) imaging, multiphoton microscopy (MPM), and various image processing tools, the whole process of 3D tissue characterization could be achieved in a non-invasive, precise, and quantitative way. A proof-of-concept AFS with blended fibers made of polycaprolactone and porcine gelatin was used to demonstrate the feasibility of implementing such a strategy. The data indicated that, in terms of scaffold characterization, the proposed MPM method was capable of measuring the porosity of homogenous scaffolds precisely from deconvolved 3D images. Furthermore, the method could also be used to illustrate the orientation of the aligned nanofibers. Next, when SH-SY5Y neurons were cultured on the AFS, the MPM imaging was capable of evaluating the cell viability ratio, cell-localization in AFS, and neurite outgrowth. This provided guidance for selecting the alignment method for AFS functional recovery. Lastly, when employing this non-invasive imaging-based characterization method, it was possible to illustrate the relationship between the alignment of collagen arrays in decellularized corneal stroma and the transparency. In summary, the proposed strategy can provide some essential scaffold/tissue properties (such as alignment of fiber, porosity of scaffold, and cell viability ratio) quantitatively and non-invasively, which will help both scaffold processing design and characterization.

Mesoporous silica nanoparticles (MSNPs) are of interest as effective drug carriers because of their controllable physical properties and biomedical compatibilities. A number of different MSNPs have been assessed for their suitability as intracellular nano-scale carriers of chemotherapeutics and siRNAs. Four morphologically different MSNPs were synthesised after optimisation of existing protocols. The MSNPs were characterised with regards to size, porosity, surface area, surface charge, cytotoxicity and biodegradability. Their suitability as drug carrier in vivo was examined in terms of cargo loading, ability to be endocytosed by cells and take ad-vantage of the Enhanced Permeability and Retention effect. The loading and unloading profiles of two model compounds and a potential chemotherapeutic agent LY294002 were investigated. The release behaviours of the cargoes were altered by modifying the particle surface with polymeric capping agents. In addition, the particles were capped with pH-sensitive molecules, and the release behaviour in low pH was assessed since tumours are known to have an acidic microenvironment. The physiological function of LY294002 on selected cancer cell lines was also studied. LY294002 was shown to affect the proliferation, survival, and metabolism of selected cells under different oxidative conditions. The effect differed when cells were under oxi-dative stress and/or glucose stress. Cell viability was also compromised after treatment with LY294002 loaded MSNPs. The sensitivity to each LY294002 loaded MSNP differed between cell lines. Engulfment and cell motility 1 (ELMO1) - targeted siRNA was also delivered using MSNPs to two distinct rhabdomyosarcoma lines. Significant knock-down of the ELMO1 gene was shown, illustrating that MSNPs could be efficient transfection agents for siRNA. In particular, the two MSNP candidates were shown to be significantly better than a current commercial product. A co-delivery system for LY294002 and ELMO1-targeted siRNA was established. Cell viability and ELMO1 expression were both suppressed after treatment with the co-delivery system.

Currently, 30-35% of the global population uses the Internet. Furthermore, there is a rapidly increasing number of non-English language internet users, accompanied by an also increasing amount of unstructured text online. One area replete with underexploited online text is the Arabic medical domain, and one method that can be used to extract valuable data from Arabic medical texts is Named Entity Recognition (NER). NER is the process by which a system can automatically detect and categorise Named Entities (NE). NER has numerous applications in many domains, and medical texts are no exception. NER applied to the medical domain could assist in detection of patterns in medical records, allowing doctors to make better diagnoses and treatment decisions, enabling medical staff to quickly assess a patient's records and ensuring that patients are informed about their data, as just a few examples. However, all these applications would require a very high level of accuracy. To improve the accuracy of NER in this domain, new approaches need to be developed that are tailored to the types of named entities to be extracted and categorised. In an effort to solve this problem, this research applied Bayesian Belief Networks (BBN) to the process. BBN, a probabilistic model for prediction of random variables and their dependencies, can be used to detect and predict entities. The aim of this research is to apply BBN to the NER task to extract relevant medical entities such as disease names, symptoms, treatment methods, and diagnosis methods from modern Arabic texts in the medical domain. To achieve this aim, a new corpus related to the medical domain has been built and annotated. Our BBN approach achieved a 96.60% precision, 90.79% recall, and 93.60% F-measure for the disease entity, while for the treatment method entity, it achieved 69.33%, 70.99%, and 70.15% for precision, recall, and F-measure, respectively. For the diagnosis method and symptom categories, our system achieved 84.91% and 71.34%, respectively, for precision, 53.36% and 49.34%, respectively, for recall, and 65.53% and 58.33%, for F-measure, respectively. Our BBN strategy achieved good accuracy for NEs in the categories of disease and treatment method. However, the average word length of the other two NE categories observed, diagnosis method and symptom, may have had a negative effect on their accuracy. Overall, the application of BBN to Arabic medical NER is successful, but more development is needed to improve accuracy to a standard at which the results can be applied to real medical systems.

Osteoarthritis (OA) is a debilitating disease characterised by degradation of the articular cartilage and changes in the subchondral bone. Presently the gold standard treatment for OA is total joint replacement using metal, ceramic and non-degradable polymer materials. Tissue engineering using novel bioresorbable biomaterials has the potential to stimulate regeneration of bone and cartilage for early stage intervention in OA suffers. This thesis investigates the synthesis of hydroxyapatite nanoparticles (HAp) and techniques to generate poly (lactic acid) (PLA) HAp nanocomposites. The effect of the synthesised HAp on isolated OA donor derived human mesenchymal stem cells (hMSC) was investigated in both 2D and 3D culture conditions. A highly controllable sol-gel synthesis method demonstrated control over HAp morphology and composition, with modification of titration rate, addition methodology and reaction pH. Two novel nanocomposite fabrication techniques were developed and characterised with transmission electron microscopy (TEM) demonstrating HAp dispersion at the nanoscale throughout PLA. Dip-coated HAp PLA and fibrin substrates were fabricated and demonstrated maintenance of hMSC adherence, proliferation and osteogenesis on 2D substrates. Investigations into fibrin encapsulated hMSC illustrated HAp uptake within the cell following 24 hours incubation. Further studies examining fibrin/HAp encapsulated hMSC showed increased osteogenic gene expression, peripheral matrix deposition and mineralisation following 21 days in culture. 3D printed PLA constructs infused with fibrin and fibrin/HAp encapsulated hMSC demonstrated significant osteogenic gene expression differences at day 21. However, these data were variable between cell isolations from different patients further illustrating hMSC heterogeneity and hMSC donor–donor variability in-vitro.

Brain computer interface (BCI) is a paradigm that offers an alternative communication channel between neural activity generated in the brain and the user’s external environment. BCI decodes the brain activity obtained from an electroencephalogram (EEG) signal and convert this information to a sensible output such as commands to control and communicate with the augmentative and assistive devices. Nevertheless, the majority of the existing BCI system associates with healthy subjects operate based on a combination of multiple limbs and bounded by capability of low dimensional control. Besides that, the acquired results also are not an appropriate platform to infer with the neurologically impaired patients (e.g. spinal cord injury patients). This is probably healthy subjects have full control over their limbs and their EEG signatures show a different pattern. On the other hand, neurologically impaired patients have limited access/control over their limbs and the EEG signatures are affected by the side effects of the prescribed medication, deafferentation and cortical reorganization of brain regions as a function of duration, level and type of disease. This study focuses on the feasibility of developing a multi degree of freedom control BCI system using imagination and intention of movement of a single limb for spinal cord injury (SCI) patients. A pilot study has been conducted on eleven healthy subjects to examine the feasibility of the proposed experimental protocol to record data for implementing the same procedure on SCI patients. In the present study, eighteen SCI patients from Queen Elizabeth National Spinal Injury Unit of the Queen Elizabeth University Hospital voluntarily participated as subjects. The participating subjects have performed and imagined performing right wrist movement towards four centre out directions using a custom made manipulandum triggered by a visual cue whilst EEG, electromyography (EMG) and movement signals are recorded simultaneously through NeuroScanTM Synamp system and CED 1401 (Cambridge Electronic Design). The EEG signal was analysed using signal processing and statistical analysis method. Our findings indicate the detection of Bereitschaft potential 500ms before onset of movement and 500ms after onset of the visual cue. Additionally, there are statistical differences in the relative power within vi the EEG signal rhythm components namely, delta, theta, alpha beta and gamma bands during imagination and intention of movement towards the four different directions. The significant changes of the estimated relative power of EEG components were extracted as features associated with direction. The features then were normalised, cross validated and dimensionality reduced before being classified using k nearest neighbour (k-NN), fuzzy k nearest neighbour (FKNN) and quadratic discriminant analysis (QDA) classifier. The single trial classification results for motor imagery and motor task by k-NN, FKNN and QDA classifier dwell within the range of 52.31%-94.14% and 52.20%-96.51%, respectively. These findings proved that it is possible to develop a functional multi degree of freedom BCI system that employs imagination/intention of movement using a single limb for the SCI population. On top of that the developed BCI system and classification also required no subject training at all.

Brain-computer interfaces (BCI) provide an alternative communication channel which does not rely on the brain's normal output pathway between patients suffering from neuromuscular diseases and their external environment. BCI requires at least one brain signal as input in order to interpret the intent of the user. Non-invasive electroencephalography (EEG) is the most common and favourite method for acquiring brain signals. In the last two decades, several EEG based BCIs have been developed to help these patients. The brain signals which can be recorded in EEG and used as the input for BCIs include motor sensory rhythm, slow cortical potential, P300 and steady-state visual evoked potential (SSVEP). Compared to the other EEG based BCI paradigms, SSVEP based BCI has the advantage of high information transfer rate, high detection rate, less user training time required and commands scalability. Furthermore, SSVEP based BCI is normally operated in the self paced mode which is more intuitive and practical for real world applications. Recently, SSVEP based BCIs have attracted great attention in the field of BCI research. While most SSVEP BCI studies focus on the improvement of signal detection and classification accuracy, there is a need to bridge the gap between BCI research and practice in the real world. SSVEP based BCI requires an external visual stimulator to elicit SSVEP response. Currently, for most SSVEP based BCIs the viewing distances between the visual stimulator and the users are less than 100cm, limiting the usability and flexibility of BCI and its potential applications and users. This study proposes a novel distance adaptable SSVEP BCI paradigm which allows its users to operate the system from a range of viewing distances between the user and the visual stimulator. Unlike the conventional SSVEP BCI where users can only operate the system when they are sitting in front of the visual stimulator at a fixed distance which is normally less than 100cm, in our proposed system, users can operate the BCI at any viewing distance within the range in this proposed BCI. It is hoped that the proposed BCI system can improve the usability and the flexibility of BCI and also broaden the range of potential applications and users. For example, it can be used by older people with degenerating mobility or by patients with impaired mobility in the care environment to support their independence. Moreover, it can also be used by healthy people in a smart home or for a game control environment. The primary goal of the present study is to investigate the feasibility of the proposed distance adaptable SSVEP based BCI. This study first investigates the impact of the viewing distance on SSVEP response and compensates the deteriorated SSVEP resulting from the viewing distance by changing the intensities of the visual stimuli. 10 healthy subjects participate in the experiment to assess the feasibility of the distance adaptable SSVEP based BCI. The feasibility of the system is evaluated by the classification performance of off-line experiments at different viewing distances. The classification accuracies of the proposed BCI are examined by different EEG time window lengths, number of SSVEP harmonics and the number of recording electrodes employed. This study also investigates the sources of deterioration of SSVEP detection in BCI setup and proposes an electrode ranking method to select the recording electrodes for the implementation of the real time on line system. The experimental results demonstrate that a distance adaptable SSVEP BCI is achievable and that electrodes chosen by the proposed electrode ranking method outperform electrodes chosen by random selection in classification performance.

Substantial amount of research in home-use health monitoring techniques has emerged given growing global health awareness and ageing population in recent decades. These sensor-driven home-use healthcare applications encourage patient involvement at home during daytime activities and nighttime sleep, effectively help assess patients conditions away from clinics and hospitals, and significantly reduce the number of infirmary visits. However, there are two main issues in current wearable/remote sensor-based home-use health monitoring applications: 1) portable human motion analysis systems that are commercially available still require substantial amount of manual effort to process the measurements, which is time consuming and thus impractical for long-term home-use health monitoring, and 2) current sleep-related health monitoring applications are intrusive to the body, limited to measuring the respiration rate and sleep duration, or not clinically validated to demonstrate their efficacy. In this dissertation, we overcome the drawbacks of current health monitoring systems as follows. For lower limb motion analysis, we propose an alternative to state of the art optical motion analysis systems, cost-effective and portable, single-camera system. For upper limb motion analysis, we track all relevant body joints simultaneously, and classify the post-stroke recovery levels based on features extracted from the tracked body-joint trajectories. For abnormal respiratory event detection during sleep, we propose to record video and audio of a patient using a depth camera during his/her sleep, and extract relevant features to train a classifier for detection of the abnormal respiratory events scored manually by a scientific officer based on data collected by a clinical-use sleeping device The main contribution of this dissertation lies in proposing new application-driven algorithms for advancing cost-effective human limb motion analysis and sleep monitoring healthcare techniques, including an autonomous detection scheme for finding the initial and final frames that are of interest for video analysis, a single marker tracking scheme that is based on the Kalman filter and Structural Similarity image quality assessment,an autonomous gait event detection scheme that is based on the features of the relative positions of the markers, a scheme classification of the post-stroke recovery level by minimization of graph total variation with graph-based signal processing, an alternating-frame depth video coding scheme, a depth video temporal denoising scheme using a motion vector graph smoothness prior, and a dual-ellipse model that can efficiently track the torso motion during a person is sleeping. Experimental results show that, both the autonomous frame-of-interest detection and gait event detection show high detections rates. The validation of tracking in terms of the knee angle, shoulder movement, trunk tilt and elbow movement with a gold standard optical motion analysis system shows R-squared value larger than 0.95. The graph-based classification scheme has the potential to accurately classify participants into different stroke groups. Our depth video coding scheme outperforms a competitor that records only the 8 most significant bits. Our temporal denoising scheme reduces the flickering effect without ever-smoothing. Finally, our trained classifiers can deduce respiratory events with high accuracy. Overall, our proposed limb motion analysis system offers an alternative,inexpensive and convenient solution for clinical gait and upper limb motion analysis,and our proposed sleep monitoring system can reliably detect abnormal respiratory events using our extracted video and audio features.

This thesis investigates the use of Volatile Organic Compounds (VOCs) in disease diagnosis and monitoring. VOCs may be found in the human body, in exhaled breath, faecal matter, urine, and skin. Analysis of the volatile profile produced in the human body can provide an indicator of metabolic status, allowing the screening and monitoring of different diseases and conditions, non-invasively and painlessly. In this thesis a range of highly sensitive analytical techniques have been adopted to measure such VOCs and demonstrate that such monitoring may be used as a disease diagnostic. For example breath samples may be analysed and calibrated against gas-phase standards prepared under physiologically representative concentrations as a tool for non-invasive disease monitoring, e.g. type 2 diabetes. Detailed faecal headspace analyses of two different mouse models of type 2 diabetes (Cushing´s mice and Afmid) were made. The mouse model of Cushing’s syndrome develop excessive circulating glucocorticoid concentrations, which are associated with obesity, hyperglycaemia and insulin resistance. The Afmid knockout mice suffer inactivation of Afmid genes, which in part regulates many functions including pancreatic secretion. These mice show impaired glucose tolerance. The gut microbiota of diabetic mice appear to have a different composition when compared to wild-type littermates, i.e. significantly increased levels of short-chain fatty acids (SCFAs), ketones, alcohols and aldehydes were found in the faecal headspace of diabetic mice, and a possible link between gut microbiota and type 2 diabetes is demonstrated. The use of VOCs as a screening tool of colorectal cancer was also explored. The current screening tools show lack of sensitivity and specificity for the screening of the disease. The volatile faecal profile of patients with colorectal cancer was investigated, and sulphide compounds, including hydrogen sulphide (H2S) are shown to have potential as biomarkers for screening of colorectal cancer.