The effect of data compression on image quality in high-resolution digital chest radiography.

Abstract

To efficiently use medical imaging network resources, such as data storage and communication link, data compression is necessary. Irreversible data compression methods have been proposed to reduce the data storage and communication requirements of digital imaging systems. This dissertation reports on the relationship between the degree of compression and the detectability of induced error using radiologic observers. A new irreversible compression method which incorporates both the discrete cosine transform (DCT), a unitary transform, and reversible coding methods was developed. In general, the error produced by compression and decompression increases as an algorithm's compression ratio is increased. The study was conducted with chest radiographs selected and ranked for subtle evidence of interstitial disease, pulmonary nodules, or pneumothoraces. Images were digitized at 86 microns (4K $\times$ 5K) and 2K $\times$ 2K regions extracted. Data was compressed using the improved full-frame DCT method at compression ratios varying between 6:1 and 60:1. The decompressed image was reprinted next to the original uncompressed image in a randomized order with a laser printer. A paired-film observer study with multiple experienced radiologists was used to test whether induced errors were detectable. The results of this study defined conditions for which errors associated with the irreversible compression method cannot be detected by radiologic observers. Compression with ratios under 6:1 does not introduce any detectable degradation. The compression ratio threshold dividing the detectable and undetectable degradation is 10:1. Notably, the study was designed to test whether observers could detect the induced error and was not designed to test whether detected errors alter diagnostic decisions. The use of a film digitizer and a film printer which could replicate all of the contrast and detail in the original radiograph made the results of this study insensitive to instrument performance and dependent on radiographic image quality. An objective measure, power spectrum of the error image, was used to characterized the error introduced in the compression process; the measure was found to be closely related to the subjective views of the observer study. It describes the magnitude of errors at difference frequencies.