Acute skin irritation is the reversible inflammatory response of the epidermis to a topically applied irritant substance. A tissue engineered model of the epidermis is used to test chemicals. The degree of development of the model needs to be carefully judged in order to get the correct proportion of proliferating through to differentiated phenotypes for normal function. This judgement typically necessitates over sensitive models with an underdeveloped barrier functionality, as opposed to an insensitive model due to terminal differentiation and low numbers of basal keratinocytes. It has been reported that the lack of proliferative epidermal cells in cultures may be due to the absence of fibroblasts. Paracrine signalling in response to potential irritants is required for propagating an acute inflammatory response. The aim of this thesis is to develop a skin model using a Three Dimensional scaffold that accurately mimics the micro-environment at the DEJ, for supporting keratinocyte and fibroblast self-organisation. We hypothesise that it takes a full thickness skin model with a complete cascade of inflammatory stimuli and cytokine signalling to provide a real indication of irritation. Initial studies focused on Alvetex® (Reinnervate Ltd.), a highly porous polystyrene scaffold, with the aim of developing a skin model using the immortalised cell line HaCaT (human adult low calcium high temperature) keratinocytes or NhKs, in co-culture with dermal fibroblasts. Skin models using electrospun biodegradable polymer scaffolds made of Poly L-lactide (PLLA) and a Poly L-lactide/Polyhydroxybutyrate-co-hydroxyvalerate/Poly L-lactide (PLLA/PHBV/PLLA) composites were then developed. Issues with achieving epidermal-dermal separation in the Alvetex® scaffold due to keratinocyte entrapment lead to an Alvetex®-PHBV Bilayer. Concentration of the SDS needed to illicit an irritant response was deduced at 2D to be 0.1-0.15mM, 3D submerged to be 0.33-0.5mM and for 3D air-liquid models were at best unaffected by 8mM SDS with a Bilayer scaffold of PHBV-PLLA.

This work concerns the detection of relationships between key information in biomedical publications, such as treatments for diseases or side-effects of drugs. Given a sentence containing some medical concepts the goal is to determine their relationship to each other. Supervised machine learning methods are a very popular way to address this problem and often provide reliable results. Those methods require manually labelled examples to extract characteristics of particular relationships in order to detect similar information in unlabelled data. However, manually labelled data is not always available and its generation is time consuming and expensive. The main objective of this thesis is the exploration of distant supervision, a method which generates those labelled examples automatically using prior knowledge to detect relationships between key facts. First, relation extraction using a limited amount of training data is explored to detect adverse-drug effects in natural language. Then, work focuses on automatically labelling data using a large biomedical knowledge base, the Unified Medical Language System (UMLS). The effectiveness of a popular evaluation method that does not require manually labelled data is examined in more detail. The main goal is the investigation of whether UMLS is suitable to be used to label data automatically so as to detect similar information in natural language. Finally, a method to reduce falsely labelled instances in the automatically generated data is presented and found to improve the detection of relationships.

The main deficiency in the modern medical flexible endoscope is that it confers limited independent instrument movement. Endoscopic multitasking platforms have been designed to overcome this deficiency. These platforms deliver independent instrument movement through the use of traction cables. However, traction cable confers limited instrument movement precision in a flexible endoscope. Improved instrument control could enable more clinicians to perform advanced endoscopic techniques. The aim of this PhD is to design, build and test a novel robotic endoscopic platform which uses micro motors. The use of inbuilt micro motor for instrument actuation in a flexible endoscope has never been described before in the literature. With an improved platform, more patients could benefit from effective and safe minimally invasive therapy. The proposed platform uses in-built motors located at the endoscope tip, endoscope handle and an external unit which together exert forces on any suitable flexible instruments in order to produce up to five degree of independent instrument movement. The design and construction of the twin channel endoscopic platform is performed using computer aided design and rapid prototyping metal printing technology. The key to the successful development of the platform is the development of a novel four-piece linkage mechanism located at the tip of the endoscope, which is capable of guiding instrument movement with two degrees of freedom. Bench top analysis consisted of three parts. Firstly, kinematic performance of the prototype is compared with the predicted performance based on computer simulation. Secondly, force analysis is performed using a traction force gauge. Thirdly, an upper gastrointestinal phantom is used to test the ability of the novel platform in accessing the human upper gastrointestinal tract. A basic functional prototype of the novel platform is constructed. The motors can be controlled with a standard game controller. Kinematic analysis demonstrated that the prototype range of movement is similar to that of the computer simulation model. Force analysis revealed that the prototype is capable of generating a force of 0.45 – 5.94N dependent on the direction of instrument movement. This compares favourably with the conventional endoscope, which is capable of generating 0.4N force. Currently, the prototype is designed to be manufactured using metal printing technology. Therefore, the prototype has been designed with especially thick parts in order to overcome the limitations of this manufacturing technology. The dimensions of the manufactured prototype are 25mm (horizontal), 16.4 mm (vertical) and 61 mm (length). Although this dimension is similar to other published endoscopic multitasking platform and existing endoscopic ultrasound probes, the prototype could not negotiate beyond the oropharynx of the phantom.

We have recently developed a nanocomposite based on polyhedral oligomeric silsesquixoane for medical device application for particular use in a microvascular network. Using spectroscopic analyses, we characterised the polymer before subjecting it to both in vitro and in vivo degradation. In the same setting, the foreign- body reaction of this polymer was determined. An in vitro feasibility study as to the efficacy of neo-endothelialisation was then performed. Next, the anti-thrombogenic nature of the nanocomposite was elicited. Based on the favourable outcomes of these experiments, nanocomposite microvessels were designed and fabricated to form the components of a microvascular network. Our results indicate that silsesquioxane is an optimal building material for microvascular grafts. This therefore constitutes the foundation on which we intend to construct an in vitro artificial capillary bed model.

The aim of joint arthroplasty is to reduce pain and improve the range of motion and functionality in joints affected by diseases such as osteoarthritis and rheumatoid arthritis. Data recorded in National Joint Registries offers the clinical perspective in relation to prosthesis failures; however, this does not explain why a prosthesis has failed. Surgeons performing revision surgery for different implants often report similar findings, despite designs of prostheses and the natural joints having numerous differences, including anatomy, loading and range of movement. The underlying factor in the majority of cases of implant failure is complications arising as a result of wear debris. To understand the failures of artificial joints, a series of studies were performed examining hip prostheses in pre-clinical and post-clinical scenarios and finger prostheses in a post-clinical scenario. The pre-clinical studies focussed on areas including: the effect of acetabular shell deformation; and validating a method to measure volumetric wear from femoral stem trunnions. The deformation studies included an investigation of how bone strength influenced deformation. The post-clinical studies involved analysing retrieved finger and hip prostheses, to quantify the damage surfaces had sustained in vivo. Analysis of the finger prostheses involved the use of a non-contacting surface profilometer, to determine the surface roughness, whilst for the hip prostheses a coordinate measuring machine was used to quantify the volumetric wear. The deformation studies found that the maximum deformation was 340 μm, which could be sufficient to disrupt the assembly process of modular acetabular components. The strength of the bone was not found to correlate with the size of the deformation. The validation study found that the coordinate measuring machine was able to measure trunnions with a maximum error of 0.13 mm³ compared with gravimetric measurements. The ex vivo cohort of trunnions had a median wear volume of 0.14 mm³ (range 0.04 – 0.28 mm³). The first finger study analysed coated, metal-on-metal prostheses finding that prostheses had suffered extensive wear on the articulating surfaces. This was hypothesised to be due to the failure of the coating interface, resulting in a hard ii “grinding paste” that wore the articulating surfaces. The second finger study examined a cohort of explanted pyrolytic carbon prostheses. Even after use in vivo the roughness average (Ra) for the articulating surfaces was below the 50 nm specified by British Standards as the maximum Ra for orthopaedic implants manufactured from metal or ceramic.

This thesis aims to verify a possible benefit lossless data compression and reduction techniques can bring to a wearable and wireless biomedical device, which is anticipated to be system power saving. A wireless transceiver is one of the main contributors to the system power of a wireless biomedical sensing device, and reducing the data transmitted by the transceiver with a minimum hardware cost can therefore help to save the power. This thesis is going to investigate the impact of the data compression and reduction on the system power of a wearable and wireless biomedical device and trying to find a proper compression technique that can achieve power saving of the device. The thesis first examines some widely used lossy and lossless data compression and reduction techniques for biomedical data, especially EEG data. Then it introduces a novel lossless biomedical data compression technique designed for this research called Log2 sub-band encoding. The thesis then moves on to the biomedical data compression evaluation of the Log2 sub-band encoding and an existing 2-stage technique consisting of the DPCM and the Huffman encoding. The next part of this thesis explores the signal classification potential of the Log2 sub-band encoding. It was found that some of the signal features extracted as a by-product during the Log2 sub-band encoding process could be used to detect certain signal events like epileptic seizures, with a proper method. The final section of the thesis focuses on the power analysis of the hardware implementation of two compression techniques referred to earlier, as well as the system power analysis. The results show that the Log2 sub-band is comparable and even superior to the 2-stage technique in terms of data compression and power performance. The system power requirement of an EEG signal recorder that has the Log2 sub-band implemented is significantly reduced.

Over the years, the rapid expansion in imaging techniques and equipments has driven the demand for more automation in handling large medical and biological data sets. A wealth of approaches have been suggested as optimal solutions for their respective imaging types. These solutions span various image resolutions, modalities and contrast (staining) mechanisms. Few approaches generalise well across multiple image types, contrasts or resolution. This thesis proposes an automated pixel-level framework that addresses 2D, 2D+t and 3D structural segmentation in a more generalizable manner, yet has enough adaptability to address a number of specific image modalities, spanning retinal funduscopy, sequential fluorescein angiography and two-photon microscopy. The pixel-level segmentation scheme involves: i ) constructing a phase-invariant orientation field of the local spatial neighbourhood; ii ) combining local feature maps with intensity-based measures in a structural patch context; iii ) using a complex supervised learning process to interpret the combination of all the elements in the patch in order to reach a classification decision. This has the advantage of transferability from retinal blood vessels in 2D to neural structures in 3D. To process the temporal components in non-standard 2D+t retinal angiography sequences, we first introduce a co-registration procedure: at the pairwise level, we combine projective RANSAC with a quadratic homography transformation to map the coordinate systems between any two frames. At the joint level, we construct a hierarchical approach in order for each individual frame to be registered to the global reference intra- and inter- sequence(s). We then take a non-training approach that searches in both the spatial neighbourhood of each pixel and the filter output across varying scales to locate and link microvascular centrelines to (sub-) pixel accuracy. In essence, this \link while extract" piece-wise segmentation approach combines the local phase-invariant orientation field information with additional local phase estimates to obtain a soft classification of the centreline (sub-) pixel locations. Unlike retinal segmentation problems where vasculature is the main focus, 3D neural segmentation requires additional exibility, allowing a variety of structures of anatomical importance yet with different geometric properties to be differentiated both from the background and against other structures. Notably, cellular structures, such as Purkinje cells, neural dendrites and interneurons, all display certain elongation along their medial axes, yet each class has a characteristic shape captured by an orientation field that distinguishes it from other structures. To take this into consideration, we introduce a 5D orientation mapping to capture these orientation properties. This mapping is incorporated into the local feature map description prior to a learning machine. Extensive performance evaluations and validation of each of the techniques presented in this thesis is carried out. For retinal fundus images, we compute Receiver Operating Characteristic (ROC) curves on existing public databases (DRIVE & STARE) to assess and compare our algorithms with other benchmark methods. For 2D+t retinal angiography sequences, we compute the error metrics ("Centreline Error") of our scheme with other benchmark methods. For microscopic cortical data stacks, we present segmentation results on both surrogate data with known ground-truth and experimental rat cerebellar cortex two-photon microscopic tissue stacks.

Many tissues have the fascinating ability to self-heal or remodel when experiencing stresses or trauma. However, above a critical defect size, our body cannot regenerate by itself leading to the formation of non-functional scar tissue. Biomaterials can be used to provide a temporary template for the damaged tissues, facilitating their full regeneration. However, the synthesis and use of such materials must adequately respond to the specific needs of the tissue targeted and therefore fulfil distinctive criteria. It is particularly true for the reconstruction of bone tissue, still lacking of an ideal synthetic biomaterials template. In this thesis, a biomimetic approach was developed to synthesise an ideal im- plant for the regeneration of hard tissues. A bottom-up strategy was used based on the sol-gel process where inorganic/organic hybrid co-networks were fabricated. To do so, bespoke polymers were synthesised containing alkoxysilane precursors which can be used to covalently bond to the growing silica network during the sol-gel process. A particular attention was brought to polymers with a high degree of cross-linking in particular homopolymers of 3-(trimethoxysilyl)propyl methacrylate and N- [3-(trimethoxysilyl)propyl] acrylamide. Models were developed and applied to experimental data to get a better insight on how these polymers affect the sol-gel process as well as the structure and properties of their resulting hybrids, as a function of the inorganic to organic ratio, molecular weight, polydispersity and synthesis methods. A good understanding of these materials is crucial to improve their properties, progressing towards an ideal implant. Hybrids were found to outperform their pure inorganic equivalent in terms of me- chanical properties, nucleation of bone like minerals, cell attachment and proliferation, presenting a huge potential for the regeneration of hard tissue.

The auditory ribbon synapse is highly specialised to regulate the release of glutamate from IHCs and generate action potentials in auditory afferent fibres in response to small and graded changes in the receptor potential of IHCs. This is essential for maintaining the fidelity of auditory stimuli over wide range of frequencies and intensities. This study was aimed at identifying possible novel molecular regulators for the development and function of auditory ribbon synapses. I have used patch-clamp electrophysiology to record calcium currents and changes in membrane capacitance from IHCs to monitor vesicle fusion at the auditory ribbon synapse. The ubiquitously expressed calcium-sensor for vesicle priming, DOC2B was considered to be a potential priming factor at the auditory ribbon synapse. I have found that it is not involved in fast, calcium-dependent exocytosis of synaptic vesicles at the auditory ribbon synapse. The inositol 5’-phosphatase SynJ2 is required for the survival of auditory hair cells and is thought to regulate clathrin-mediated endocytosis. My results show that IHCs from SynJ2 KO mice display normal endocytic responses after exocytotic events and normal replenishment of the vesicle pools. Therefore, SynJ2 is not required for endocytosis, or vesicle recycling in IHCs. Connexins 26 and 30 are subunits of heteromeric gap-junctions and hemichannels in supporting cells of the cochlea. Connexin mutations cause over 50% of cases of non-syndromic hereditary deafness in humans. IHCs from mice with severely reduced connexin expression display larger calcium currents and smaller exocytotic responses. Therefore expression of connexins in the cochlea is essential for presynaptic function at the auditory ribbon synapse. Finally I have found that the transcriptional co-activator Wbp2 is not required for presynaptic function of mature IHCs, therefore Wbp2 is not involved in the transcription of key presynaptic molecules in IHCs.

There is growing clinical interest in non-invasive techniques to diagnose disease – these include breath analysis, cardiovascular and imaging. The latter consisting of both Thermal infrared (TIR) imaging where the distribution of body heat at the skin can be measured, and spectral imaging where pigments within the skin can be detected. Recent developments in imaging technology are allowing both TIR and spectral imagers to achieve good spatial resolution. Such techniques have been implemented at the Leicester Royal Infirmary (LRI) within the Diagnostics Development Unit (DDU). The majority of studies to date using these techniques have used a controlled laboratory environment. In this work the utility and practicality of both TIR imaging and hyperspectral imaging in a real clinical environment were investigated. Based upon previous work appropriate measurement and calibration protocols were developed. The clinical utility of using the available DDU imagers and protocols was investigated via ethically approved TIR studies of fever and sepsis patients and hyperspectral studies of kidney, liver and skin lesions patients. This work was conducted in the Accident and Emergency Department at the LRI and other appropriate clinical units within the University Hospitals of Leicester NHS trust. TIR imaging studies of fever, sepsis patients showed an ability to detect significant temperature abnormalities both in terms of actual values and also in terms of their spatial distribution. This includes detection of mottling pattern which may be diagnostic of progression towards sepsis. Hyperspectral imaging showed some signatures associated with liver disease, particularly within the sclera of the eye. The possible confounding effect of tanning in hyperspectral imaging of the skin was investigated. It is clear that further development is required for practical utilisation of the hyperspectral technique within the clinical environment. The main output of this work has been to develop protocols for use of the two techniques.