Microwave thermography has been used for the objective assessment of inflammation in the knee joints and wrist and finger joints of patients suffering with rheumatoid arthritis by comparison with similar information obtained from a control group of subjects. Combined microwave and thermal modelling has been used to estimate the effective blood supply to the anterior intra-articular region of the patella, and the perfusion of the quadriceps muscle in both groups. 2-D numerical modelling was compared with results obtained using 1-D modelling. Microwave thermography has also been used for the detection of breast cancer. However, problems such as high false positive detection rates have occurred due to natural cyclical breast temperature changes. The thermal behaviour of the normal breast throughout the menstrual cycle has been investigated and it is shown that microwave thermography is capable of detecting temperature variations in the female breast corresponding to the ovulatory and luteal phase of the menstrual cycle. Combined microwave and thermal modelling estimated the effective perfusion of the normal breast to be in the range 0.2 - ˜ 2 kg m-3s-1. This is consistent with previous work. Microwave thermography is a quick, simple technique which clinicians can easily use. It is non-invasive, passive and causes the patient no distress. By using combined microwave and thermal modelling it is possible to estimate tissue blood perfusions and water contents and compare them with expected values. The technique has many potential applications and will hopefully find a secure niche in clinical medicine.

Osteoporosis is a disease characterised by reduced bone mass and altered microarchitecture leading to an increased risk of fracture. The consequences of osteoporosis include reduced quality of life and pain, associated with fractures. Its financial burden on health services are significant. Characterisation of osteoporosis using imaging techniques is therefore important. Peripheral Quantitative Computed Tomography (pQCT) is a cross-sectional imaging method which is used to scan bones in the appendicular skeleton. pQCT imaging may be particularly useful in clinical groups where changes in bone mineral density (BMD) and structure are known to occur in the limbs. Two such groups are patients following spinal cord injury (SCI) or anterior cruciate ligament (ACL) injury. Aims. This project aimed to develop analysis techniques to characterise bone in pQCT images. Their purpose was to describe localised changes within pQCT images of the bone, as opposed to the standard global measurements. Methods. Fully automated segmentation and registration software was developed and tested followed by two independent processing algorithms. The first generates spatial maps to characterise local changes in BMD. This is achieved using both quadrant analysis software and a voxel-based approach, the latter comparing pairs of images and generating a voxel-by-voxel ΔBMD map of changes in BMD. The second processing algorithm uses morphological granulometries to investigate the bone microarchitecture. Results. Evaluation of these image analysis methods was carried out using two clinical studies. The first investigates acute longitudinal changes in the distal tibia (DT) and distal femur (DF) post-motor-complete-SCI using pQCT. Images from 15 subjects (13M, 2F) with a mean age of 36y±19y, were acquired at 4-monthly intervals during the first year post-injury. The second comprises of ACL injury subjects, with imaging of the injured and contralateral proximal tibia (PT) and distal femur before (n=19, 18M 1F, 30y±9y of age) and after (n=8, 8M 0F, 31y±9y of age) surgical ACL reconstruction. The software developed to automatically segment bone from surrounding structures was successful: 98% success rate for epiphyseal tibial regions, 67% success rate for the distal femur. Registration of images was then performed and the spatial analysis methods to automatically produce quadrants of trabecular bone were applied, displaying individual results graphically. The voxel based analysis method was developed, tested and applied to produce ΔBMD maps, utilising statistical inference and corrections for multiple comparisons using a false-discovery rate technique. These maps characterised localised changes in BMD between pairs of both longitudinal and contralateral images. Software was also developed to apply morphological granulometries to pQCT images, calculating global and local pattern spectrum moments. On application of the analysis methods to the longitudinal SCI images, the BMD and microarchitecture findings were observed to be disparate amongst subjects, with large variations in bone characteristics both globally and regionally. The quadrant and voxel based analysis methods provided information on longitudinal regional changes in each subject, indicating individual patterns of change. Structural analysis of bone microarchitecture using granulometries was demonstrated to have potential as a useful adjunct to BMD in identifying SCI subjects more susceptible to rapid bone loss. The analysis methods were also successfully applied to the ACL injury subjects. Following segmentation and registration, the total and trabecular BMD in the injured knee was observed to be significantly lower than that of the contralateral control knee pre-operatively for both the PT and DF (p<0.05). Post-operatively the total and trabecular BMD in the injured DF remained significantly low (p<0.05), however the PT demonstrated significantly lower BMD in the injured leg for the trabecular bone only (p<0.05). Reduced BMD in the PT post-operatively in humans is a novel observation, and indicates a benefit afforded by segmenting trabecular from cortical bone. Regional analysis using quadrants indicated some anatomical variation in bone loss within the injured limb, although it is acknowledged that these are preliminary findings which would require to be confirmed in larger studies. The voxel ΔBMD maps generally indicated global losses across the bones of the ACL injured leg both pre-operatively and post-operatively. No consistent patterns were obtained in the ΔBMD maps for these subjects, suggesting individual patterns of response to ACL injury. The structural information provided by granulometric analysis was limited for the ACL study. Conclusions. Automated software has been developed to characterise bone in pQCT images of the appendicular skeleton. It has been successfully applied to two clinical studies, facilitating localised changes in bone density to be demonstrated and descriptions of microarchitecture to be provided. The SCI subjects appear to have individualistic responses to injury, with a wide range of changes in bone density and microarchitecture observed. ACL injury patients all lost bone mass, but patterns of change were variable. The analysis methods developed to permit characterisation of bones in individual subjects, are proposed to be of value in both clinical and research domains exploring bone mass and microarchitecture, with the ultimate goals being the prediction of fracture risk and tailoring therapy for the individual.

The Frank-Starling law states that the stroke volume of a regular cardiac beat increases in response to an increase in the volume of blood filling the heart. If this law applies in atrial fibrillation (AF) as well as in sinus rhythm (SR) then cardiac function will depend on the duration of diastole in the preceding beat as well as the duration of the indexed beat. Aim: The aim of this thesis was to develop a series of tools which would allow an assessment of the changes in cardiac function from one beat to the next in AF and SR. A secondary aim was to find a means of describing rhythm in a way that reflected possible functional change. Methods: List-mode radionuclide ventriculography, RNVG, acquisitions of 373 patients in AF and a comparative group of 385 patients in SR were made. Software was written which allowed tightly defined preceding and indexed beat selection criteria to be established. Left ventricular ejection fraction (LVEF) and other functional parameters (pre-systolic volume, systolic time, the ratio of pre-systolic to end-diastolic volume, peak filling rate and first third filling fraction) were calculated for images created using different beat selection criteria based on the quartiles of beat length. Assessment used both variable and fixed time formatting and included a comparison of results achieved in the first and second half of the scan. Traditional linear measures of heart rate variability together with descriptors of the Poincar´e plot and cycle length entropy were used to describe rhythm in both AF and SR patients. Results: Substantial variation with indexed and preceding beat length was seen in both SR and AF in all the systolic parameters measured and in particular in LVEF where the standard deviation of LVEF for any one patient was found to be 8.2% in SR and 14.1% in AF. A combination of descriptors of rhythm was found to have good correlation with the range of LVEF measured. Examination of the results for LVEF in several clinical sub-groups suggests that the range of LVEF may have clinical interest. The techniques were applied in a small clinical study which considered the value of radio-frequency ablation in patients with AF and heart failure. In this study, measures of Sample entropy and the range of LVEF appeared to have prognostic value. Conclusion: A tool which allows the investigation of beat-to-beat functional variation in RNVG has been produced. It has been shown that the functional variation depending on beat selection criteria is substantial and may have clinical significance both in patients with underlying pathology and prognostically in patients undergoing radiofrequency ablation (RFA).