Automatic Readout of Contrast-Detail Curve Used in Digital Mammography Using Morphological Granulometry

Abstract

Mammography is one of the most effective methods used today for early breast cancer detection. Mammography equipment, is able to take breast radiographies, and is specially designed for it. Quality Control and Quality Assurance, provides a guarantee for the best possible image quality with an acceptable radiation dose. Threshold contrast visibility is one of several tests used in quality control, with international recognition for the acceptance of the system. This test is possible thanks to image processing of the CDMAM (Contrast-Detail Mammography) Phantom.
This paper presents a new algorithm, called MIQ (MammoIQ), for evaluating the CDMAM phantom used in digital mammography. The method consists on evaluating each cell using Mathematical Morphology and digital image processing techniques to determine the eccentric disks. Granulometry concepts are introduced for the CDMAM phantom, pattern spectrum and size distribution. The results are compared with the CDCOM software and a human observer, through the Image Quality Factor (IQF) parameter and Contrast-Detail Curve. The combined results, allow to compare different curves from several images, and to compare the performance of different systems.
Twenty-eight (28) images are analyzed. They are acquired from three different systems, and gathered into three different groups (Systems 1, 2 and 3) according to the system where they were acquired. The calculated metrics used are Contrast-Detail Curve, Size Distribution, Pattern Spectrum, and Image Quality Factor (IQF). System 2 presents a better contrast-to-detail ratio with respect to the other systems. The results obtained with the algorithm MIQ present a greater similarity to the results obtained by human observers. The presented algorithm provides non-ambiguous and fast results for expert and non-expert evaluators.