Using texture to predict diagnosis and disease from nuclear medicine lung perfusion scans: A comparison of nuclear medicine physicians to the slope of the power spectrum.
AuthorKer, Mary Virginia.
MetadataShow full item record
PublisherThe University of Arizona.
RightsCopyright © is held by the author. Digital access to this material is made possible by the University Libraries, University of Arizona. Further transmission, reproduction or presentation (such as public display or performance) of protected items is prohibited except with permission of the author.
AbstractThe lung has been satisfactorily modelled as a fractal, and change in lung structure due to disease is assumed to change the fractal dimensionality of the lung. It is hypothesized that those changes in fractal dimension affect perceptually relevant elements (perceived texture) of the lung, and therefore the fractal dimension may prove to be a predictor of diagnosis. If the fractal dimensionality reflects structure in ways more accurately reflecting changes in lung structure than can be achieved by nuclear medicine physicians, then it may also prove useful as a diagnostic tool. Fractal dimension is linearly related to the slope of the power spectrum (SPS) as plotted on log-log paper, and the SPS was used as the metric reflecting the fractal dimension. Seventy-two cases were selected that were either normal, had congestive heart failure (CHF), chronic obstructive pulmonary disease (COPD), or pulmonary embolism (PE). Five of the cases had both CHF and COPD. The lung scans from these cases were digitized, with appropriate corrections for linearization, edge artifacts, target nonuniformities and film gamma. Fast Fourier Transforms provided the power spectrum from which the SPS was calculated. Four nuclear medicine physicians read the original lung scans and rated their certainty about the presence of two texture elements, the extensiveness of disease involvement, and presence of the three diseases used (CHF, COPD, and PE). The results found the SPS to be significantly related to both texture ratings and diagnostic certainty, but inferior as a predictor of disease to either texture rating or diagnostic certainty. This study reveals the SPS to be a promising but incomplete candidate for machine-algorithm generated diagnosis.