Quantitative techniques in radioisotope imaging

The items of equipment comprising a radioisotope imaging system are briefly described, namely the Ohio-Nuclear 8100 gamma camera, the Varian 620/10-100 minicomputer, the interface and the colour T.V. display, together with the relevant software. The performance of the imaging system was analysed by line spread functions using 99mTc from which the Full Width at Half Maximum (FWHM) and the Modulation Transfer Functions (M.T.F.) were calculated. In this way spatial frequency response was determined at different distances from the "High Resolution" and "Ultra High Resolution" collimators in air. The effect on overall spatial resolution of the number of computer memory locations used for image storage was investigated by adjusting the gain of the A.D.C. in the X direction to 1, 2, 4 and 16 times normal gain and the M.T.F.'s were calculated. The calculation of the M.T.F. was done by using a program written in BASIC. A study was carried out of the application of the mini-computer interfaced to the gamma camera in improving diagnosis for static images. The simple methods of image processing, namely smoothing and spatial frequency filtering, were used. Smoothing was done on a 3 x 3 matrix (nine point smoothing) by the smoothing function suggested by Nakamura et al (1973). Spatial frequency filtering was done on a 5 x 5 matrix (twenty five point filtering) and the weighting factors were obtained from the 99mTc point source at the appropriate distances from the High Resolution collimator. A liver phantom was made with simulated lesions of various sizes, shapes and activity ratios and pictures were taken on Polaroid and recorded on magnetic tape. A comparison of Black and White Polaroid and two modes of colour T.V. display of the raw data and the effect of smoothing and smoothing plus filtering were obtained from the Receiver Operating Characteristic curves using a group of people to look at these pictures. The program for image processing and image display were written in ASSEMBLY language. The results of the line source studies show that at the surface of Ultra High Resolution collimator the FWHM is 5.8 mm whereas for the High Resolution collimator it is 6.5 mm or in terms of the M.T.P. the spatial frequencies for 10/0 response are 1.4 and 1.25 cycle/cm respectively. The study of the effect of the computer matrix on resolution proves that using a 64 x 64 matrix to record the image is inadequate. The resolution is improved by a 128 x 128 matrix and even better up to 1024 x 1024, if available. The 99mTc point source study proves that smoothing causes some loss of resolution whereas filtering restores the image closer to the original distribution but creates some arte-facts. Prom the results of the "liver" phantom studies, the percentage of true positive against false positive at the confidence level "definite" for each method of display area Polaroid 27/1, T.V. raw data, linear scale 43/2, T.V. raw data statistical scale plus or minus one standard deviation 41/3. T.V. raw data, smoothing 63/3, and T.V. raw data smoothing plus filtering 69/2. When the raw data is displayed in a statistical scale of plus or minus one standard deviation (full band equals two standard deviations), the detestability is slightly less than xiv that of raw data displayed in linear scale. However, the response of the observers to the colour television display is better than Polaroid at all levels of confidence and if either smoothing or smoothing plus filtering is applied to the unprocessed images, the detectability is improved