[The evaluation of a digital system for the automatic analysis of the density of abdominal radiograms]

This paper reports the results of the evaluation of an automatic system developed to analyze radiographic densities. This system, called RADEN, has been implemented at the Institute for Information Processing of the Italian National Research Council in Pisa. Forty-three pairs of abdominal X-ray films were obtained before and after the administration of ionic iodate contrast medium (i.i.c.m.) during urographic examinations. The 86 X-ray films were acquired and digitized using a computer-controlled optoelectronic device with a sampling step ranging 125 microns to 250 microns. The resulting images were arrays of 512 x 512 pixels, each one quantified on a byte (256 gray levels). The computing system included a high-performance personal computer equipped with a video RAM board having built-in facilities for image processing and with an optical disk to archive the images. Automatic and semi-automatic procedures were developed to segment, recognize and classify the images and to characterize three homogeneous regions--i.e., bone, water and air. The output data were compared with the area values of the air densities computed directly on the X-ray films by a radiologist. Furthermore, both data sets were compared with the scores given in a blind study by four observers. The subsequent statistic analysis showed the increase in air density areas after i.i.c.m. administration and the applicability of the implemented system to the automatic examination of abdominal radiographic densities. The results encourage to believe that the proposed approach could be employed as a first step for the development of quite a more complex system oriented to X-ray image understanding and to assisted diagnosis.