Evaluation of spatial resolution using analysis texture in BCR's breast phantom images

Mammography is the main radiographic technique for imaging the breast, being indicated for cancer diagnosis and clinical monitoring. In Brazil, the National Quality Program in Mammography (NQPM) was developed. This program provides radiation exposure reference levels and evaluates the image system quality through a visual inspection of images obtained with the breast phantom. Texture is an important characteristic that can be used to identify or describe an image. It refers to the spatial arrangement of pixel's gray levels in a given region. Texture analysis technique can be used to obtain spatial resolution results from phantom's grids present in the images. Therefore, the aim of this work is to develop texture analysis algorithms to perform an automatic quality evaluation of Brazilian College of Radiology (BCR) breast phantom images. For this study, BCR phantom images were collected from different facilities in Belo Horizonte, Brazil. The methodology consists in selecting a region of interest (ROI) and then performing image background subtraction and filtration. Thereafter, statistical values are calculated to quantify the gray level spatial distribution. Statistical parameters were used to quantify the differences between periodic models. These algorithms were developed in MATLAB and the test objects evaluations were made following the NQPM parameters. The visual and computerized evaluations were compared. The results showed that the developed algorithms can differentiate the four different grids of BCR phantom images. The use of these digital image processing techniques in the test object images allows an objective evaluation. Therefore, the developed algorithms help to improve the evaluation performed by quality assurance programme specialists. (authors)