Table of contents

The ideal observer signal to noise ratio (SNR) has been derived from statistical decision theory for all of the major medical imaging modalities. This SNR provides an absolute scale for image system performance assessment and leads to instrumentation design goals and constraints for imaging system optimisation since no observer can surpass the performance of the ideal observer. The dependence of detectable detail size on exposure or imaging time follows immediately from the analysis. A framework emerges from comparing data acquisition techniques, e.g. reconstruction from projections versus Fourier methods in NMR imaging, and time of flight positron emission tomography (TOFPET) versus conventional PET. The approach of studying the ideal observer is motivated by measurements on human observers which show that they can come close to the performance of the ideal observer, except when the image noise has negative correlations-as in images reconstructed from projections-where they suffer a small but significant penalty.

A general analysis of the signal to noise ratio SNR, of X-ray imaging with a broad spectrum is presented. The analysis indicates that the energy modulation of the signal together with its degree of matching by the energy response of the image receptor are significant determinants of the SNR for signal detection. This requires a generalisation of the interpretation of detective quantum efficiency, DQE, the transfer function appropriate from SNR, that will be dependent on the image detection or discrimination task. The generalised DQE is similar to the conventional DQE for the task of detecting radiation levels, but may differ substantially from it for the task of discriminating a lesion from its surround, particularly for signals of bone or iodine. The photon counter is shown to be inferior to the ideal detector for these tasks, but to be generally superior to the energy detecting scintillators used in conventional or digital radiography and computed tomography.

The feasibility of using heavily filtered X-ray beams to assess trabecular bone mineral content has been investigated by measuring the ratio of coherent to incoherent scattered X-rays with a high purity Ge detector. The technique uses the strong dependence of coherent scattering on the effective atomic number of the scattering medium. With an X-ray beam filtered with a high-atomic-number filter, a spectrum characterised by a sharp discontinuity at the K-absorption edge is proposed. Analysis of the spectral shape after scattering allows the coherent to Compton scattering ratio to be obtained. Theoretical and experimental results from phantom studies are presented and a comparison made between the results with X-ray beams and radionuclide sources respectively. The influence of overlying tissue thickness on the sensitivity of the measurements is demonstrated.

A stochastic lung model is proposed for aerosol deposition calculations. Airway geometry is selected randomly to reflect intrasubject variations in the human airway system. This may also be justified to take intersubject variations into account. The statistical analysis of the human airway geometry used is based on morphometric data measured at the Lovelace Inhalation Toxicology Research Institute. Average values and distributions of airway diameters and lengths, distributions of branching angles and criteria for determination of the pathway (when the alveolar region is reached) are presented. Correlations between the cross sections of tubes of succeeding generations and those between diameters and lengths of the same generation are also given.

A method for assessing regional blood flow by transient heat clearance is described. A probe at room temperature is attached to the tissue which is to be investigated and the temperature of the probe is thereby decreased. The time constants which describe the increases in tissue temperature as the temperature approaches equilibrium temperature are related to blood flow. Regional blood flow can be determined by the method of transient heat clearance without calibration.

A relationship between the structure of the sclera and its light scattering properties is proposed. The relative optical density as a function of wavelength in the infrared was measured for three sclera and a good correlation with the predicted function was found. The angular distribution of the scattered light at 632.8 nm as well as the degree of polarisation of the scattered light at zero angle was measured for three sclera; from these results the total transmission of the sclera was determined.

The optimal parameters for laser therapy of port wine stains (PWS) have been deduced from temperature calculations on two models: (i) the four-layer model; and (ii) the tube model in which two plane parallel layers, representing the epidermis and dermis, and a dermal rectangular blood vessel are considered. The calculations were performed with a vessel of average cross section 0.06 mm*0.08 mm located in the centre of the laser beam. The numerical calculations were performed by an alternate direction-implicit finite difference method. The optimal parameters were: wavelengths lambda =415, 577 and 540 nm; pulse time, 1 ms<t₁<or=10 ms; and beam radius W₁>0.1 mm. The energy densities E₁ (for t₁=1 ms) required to coagulate blood vessels down to a depth of 0.65 mm in order to establish bleaching of the PWS were 0.5, 1.6 and 2 Jcm^-2 for the different lambda respectively. The value for the argon laser ( lambda =488 and 514.5 nm) was E₁=6.5 J cm^-2 (t₁=1 ms). Because, for this pulse time, heat summation effects at the boundary of the laser beam cause no drastic increase in local temperature at optimal wavelengths, the stripe technique was again considered and compared with the separated spot technique. The Nd-YAG and CO₂ lasers prove even less selective than the argon laser. The influence of cooling the skin with water shows that only for lambda =577 nm and t₁=0.1 s is there an increase in E₁ from 2.5 to 6 J cm^-2 for which dermal damage occurs.

An analytical technique was employed to extract the primary component of dose at off-axis points from broad-beam water phantom data for a Clinac-18 Clinac-6 (100), Clinac-4 (Pb filter) and Philip's cobalt unit. The methodology presented here does not require separate off-axis HVL measurements and guarantees the accurate description of the data within the radiation field. It is shown that the off-axis reduction in the primary dose component for linacs does lead to significant (5-10%) errors in dose calculations if it is not taken into account in extended field external beam calculations. A method is described in which the off-axis variation in zero field TAR can be easily extracted from broad-beam scan data using the framework established by Cundiff et al (1973) as modified by Hanson and Berkeley (1980) for the dosimetry of large irregular fields. The technique is self-consistent and leads to parameters which have simple physical interpretations and which guarantee an accurate description of the beam.

The combination of electron filter material, thickness and position which would minimise the surface dose for a 10 MV X-ray beam from a varian Clinac 018 was investigated. It was found that a lead filter of 0.17 g cm^-2 thickness, placed below the lower collimator at a distance of 50 cm from the X-ray target, offers the greatest reduction in build-up region dose. Tin and copper filters with thicknesses of 0.19 and 0.23 g cm^-2 respectively are slightly less effective in reducing the dose in the build-up region. The maximum reduction in surface dose achieved by added filtration is only 7% for a field size of 25 cm*25 cm. Smaller improvement is achieved with smaller field sizes. In cases where blocking trays are used, the filters should be placed under the trays toward the patient's skin.