As part of the quality control of the Slovenian Breast Cancer Screening Programme (DORA), regular, preferably daily, scans of homogeneous phantom are performed at each mammography unit. Visual inspection of images is time consuming, and in addition, some iregularities, especially minor, can be overlooked. We need new methods, sensitive enough to automatically detect such irregularities.
In my thesis I introduce a method based on a simple statistical measure, skewness $g_1$, with which we can quickly detect irregularities that appear in phantom images. With this method I have discovered 13 point artefacts and 4 line artefacts that have appeared from October 2019 to March 2021 on images of mammographs that are part of the DORA programme. Some artefacts have disappeared after calibration or detector replacement, but some have stayed. For a few mammographs I made graphs of the time dependence of $P_{98\%}$ $|{g_1}|$, ${g_1}^{max}$ and ${g_1}^{min}$. I made a graph of total time dependence for Siemens and Hologic mammography systems.
The parameter $g_1$ has proven to be a reliable and robust estimator of point and line artefacts of mammography units and is a potential automated measure for quality control.
|
