Studies on x-ray physical analysis and computer-aided diagnosis of x-ray CT images [in Japanese]

The focus of this thesis was to physically analyze diagnostic x-rays by using the x-ray spectrum and to develop an automatic detection method of lymph nodes on chest x-ray CT images. This dissertation included the following five topics: (1) three issues in terms of the device, method, and data revision on the spectrum measurement of diagnostic x-rays; (2) development of a doughnut-shaped parallel ionization chamber dosimeter for monitoring the x-ray output during the measurement; (3) influence of the size of germanium crystals and different models on the spectral measurement; (4) investigation by Monte Carlo simulation of the influence of the scattered x-rays from rectangular wave chart for MTF measurements on the determined data; (5) development of a computer-aided diagnosis (CAD) system for detecting mediastinal lymph nodes on chest CT images of lung cancer patients. In the last topic, the knowledge information processing method was adapted to detect the positions of lymph nodes; anatomical information determined by referring to automatic classified slices by a genetic algorithm (GA) method and lymph node map was useful. This research shows the importance of precise measurements of physical characteristics of x-rays and the development of CAD systems for chest CT images. © 1999, American Association of Physicists in Medicine. All rights reserved.