Image processing advances show international colors

Digital radiology revealed its international nature this afternoon in a scientific session at which researchers from Brazil, India, Germany, and Poland presented papers on recent advances in image processing.

The first presentation detailed a means of characterizing lung nodules by analyzing their shape and texture on CT images.

"First, the texture is analyzed using the empirical semivariogram function," said Aristofanes Silva of the Pontifical Catholic University in Rio de Janeiro.

The shape is then examined by means of a set of measures based on the nodule's skeleton. The nodule's malignancy or benignity is determined by combining these methods.

While the number of nodules studied (31) is too small to provide definitive conclusions, such measures provide significant support for a more detailed investigation, Silva said.

Dr. Dinesh Kumar of Cochin University of Science and Technology in Kerala, India, proposed a method for automatic determination of the location and orientation of spine vertebrae in digitized spine x-rays, using mathematical morphology.

"The algorithm is based on spine morphology and hence works well even in noisy images," Kumar said.

Spine localization can be accurately determined in all cases without human intervention. Vertebrae C1 and C2 can be precisely located even when this region is unclear on the image, he said.

The proposed technique is robust for use in the initial phase of vertebrae segmentation, Kumar said.

A third paper introduced a method for localizing calcifications using the output produced by a newly developed vessel segmentation method.

"This new method generates a set of points for the vessel's centerline as well as for its border, which makes it easy to compute a display graph whose minimum can be calculated," said Dr. Stefan Wesarg of the Fraunhofer Institute for Computer Graphics in Darmstadt, Germany.

An evaluation of that minimum's position has been linked with the visually defined position of a calcification inside the vessel.

The Polish study illustrated the use of image analysis technologies for automatic detection and shape description of microcalcifications, combining wavelet, morphological, and convolution with different scale techniques.