Problems associated with processing large numbers of chest radiographs.
The use of equal probability quantization in the automated interpretation of chest radiographs for pneumoconiosis was investigated. Nineteen chest radiograph pairs were scanned to give optimal gray level resolution. The images were transformed by equal probability quantization. Film texture measurements were taken before and after quantization. A much higher individual variation for film texture measurements occurred when the image was not transformed by equal probability quantization. Film texture relied heavily on film quality in those films which had not been normalized by equal probability quantization. Classification accuracy for normal and abnormal films was inconclusive. The authors conclude that individual variability of texture measurements is reduced by preprocessing the film with equal probability quantization. This preprocessing reduces the dependency of the texture measurements on film quality.
X-ray-analysis; X-ray-diagnosis; Radiography; Diagnostic-techniques; Occupational-respiratory-disease
National Institute for Occupational Safety and Health