Comparison of Coincidence Counting and Focusing Collimators with Various Isotopes in Brain Tumour Detection

Unsharpness and contrast in digitised images

The physical quality of a digitised image is affected by non-digital processes in the imaging instrumentation and by digital processes in the computer. The spatial unsharpness depends on instrumental resolution and on the pixel size of the digital array. The image contrast depends on the properties of the imaging system and on the number of grey scales used in the digital sampling process. The possibility is examined of constructing formulae for spatial unsharpness, and for the analogous parameter in the intensity dimension, defined as threshold contrast, in terms of instrumental and digital components. Formulae are developed from simple principles and tested against the observed properties of computer simulated images. The conclusion is that the concepts and the formulae provide a basis for predicting the effects of digital processing on the total unsharpness and contrast of digitised images.

The influence of size and radiopharmaceutical concentration ratio on the detection of abnormalities in clinical radionuclide imaging

The effect of the size of an abnormality on its detectability was measured. The cylindrical abnormalities studied, with diameters ranging from 3 mm to 4 cm, were superimposed on a uniform background. To avoid artefacts produced by the imaging devices, the images were simulated by computer. The method of constant stimulus was used to determine the detectability of the abnormalities. By using a Perspex phantom filled with 99Tcm, measurements were made of the concentration ratio between abnormality and background which permitted detection of the abnormality. This allowed evaluation of the effectiveness of 99Tcm pertechnetate in the detection of small abnormalities. Assuming a concentration ratio between abnormal and normal tissue of 22.5:1, then at a depth of 50 mm in the brain a cylindrical abnormality of about 8 mm diameter was the smallest detectable under the chosen scanning conditions.

A comparison of optimum detector spatial resolution in nuclear imaging based on statistical theory and on observer performance

An expression for the expected image of a spherical tumour in a uniform background was derived in terms of background thickness and concentration of radioactivity, the tumour size, depth and uptake ratio, the gamma-ray energy and the detector response function. Three models of human observer performance for tumour detection were developed from different signal-to-noise ratio measures based on the statistical theory of detection. The optimum detector spatial resolution predicted by each model was then compared to that obtained from an observer performance study in which the subjects viewed computer-simulated scintigrams. The predictions from two of these models seem to be consistent with the results of the observer performance study. Model II involves a comparison of the counts integrated over the tumour region with the counts integrated over a background region of the same area. Model III compares the count density estimates of signal-plus-background and background obtained from application of non-uniform weighting functions to the image data.

A non-linear programming method for optimizing parallel-hole collimator design

A new method for optimizing the design of multi-aperture parallel-hole collimators for the gamma scintillation camera is presented. The method takes into account the frequency spectrum of a plane source object distribution as well as the energy of the radiation. A frequency dependent statistical figure of merit is calculated and combined with a weighted object distribution frequency spectrum to obtain an objective function which, when maximized, yields the optimum collimator design according to the chosen criteria. The optimization is performed by means of a sequential pattern search technique. The results show a positive correlation between te objective function and an experimental performance index evaluated for existing collimators. The optimal designs obtained by maximizing the objective function, under the assumption of no scatter within the source, exhibit somewhat higher sensitivity and lower resolution than the commercial low energy collimators tested. It is concluded that much of the resolution capability of very high resolution collimators is unused because of the limitation imposed by the intrinsic resolution of the detector assembly.

Computer-assisted statistical analysis of skeletal scintigrams

Forty-two consecutive skeletal scintigrams of the dorsal and lumbar spine have been statistically analysed using an Elliott 4120 Computer. Each patient had, in addition, anteroposterior and lateral radiographs of the dorsal and lumbar spine taken. All the patients have been followed up for a minimum of 18 months. In each patient the radiological, scintigraphic, and statistically analysed data have been compared and the accuracy of the results determined by the subsequent progress made by the patient. The results indicate that the computer analysis introduced both false positives and false negatives, and it is suggested that, at the moment, the computerassisted statistical analysis of the skeletal scintigram is not sufficiently reliable for clinical use.