Determination of contrast-detail curves of mammography systems

Many phantoms have been designed to study mammographic image quality. Not all of these are suitable for use with digital systems. Especially, measurement of spatial resolution with line patterns does not seem appropriate, because microcalcifications, the smallest objects visible in mammograms, are better represented by dot patterns. A phantom in which dot patterns are applied is the CDMAM phantom that was developed at our institute (Figure 1). This phantom consist of a matrix of square cells with dots of varying size and contrast. In each of the 205 cells of the matrix one dot is at the center and another is positioned in a randomly selected corner. The observer’s task is detection of the dots. Responses can be evaluated as percentages of correct detections per image. When using multiple readouts of phantom recordings the fraction of correct decisions can be determined per cell, i.e. as a function of object size and contrast. These fractions can be used to determine a contrast detail curve relating object size and contrast at some fixed detection threshold. Measurement of contrast detail curves by human readout of multiple phantom images is tedious and time consuming. Therefore, we have developed a computer program which performs an automatic readout of the phantom recordings. The use of this program allows making quick and reliable comparisons of digital mammographic imaging procedures.