Modelování textur aplikované na lékařské snímky

Abstract

This thesis presents novel descriptive multidimensional Markovian textural models applied to computer aided diagnosis in the eld of X-ray mammography. These general mathematical models, applicable in wide areas of texture modeling outside X-ray mammography as well, provide ideal visual verication using synthesis of the corresponding measured data spaces, contrary to standard discriminative models. All achieved results in the thesis are extensively benchmarked. The thesis presents two methods for breast density classication in X-ray mammography. The methods were tested on the widely known MIAS database and the state-of-the art INbreast database, with competitive results. Several methods for completely automatic mammogram texture enhancement are presented. These methods are based on the descriptive textural models developed in the thesis which automatically adapt to the analyzed X-ray texture, thus being universal for any type of input without the need of further manual tuning of specic parameters. The methods' outputs highlight regions of interest, detected as textural abnormalities. The methods provide the possibility of enhancement tuned to specic types of mammogram tissue. Hence, the enhanced mammograms can help radiologists to decrease their false negative evaluation rate. It has been shown that the algorithms work well both for small ndings, such as microcalcications, and for bigger lesions. The pseudocolour method oers a unique way of mammogram feature fusion for visual evaluation and vastly enriches the the information content of the enhanced mammogram. The results were veried also by radiologist consultants. New contrast criterion was implemented which outperforms previously published contrast criteria. The focus of our study is mammogram texture and the following search for its optimal mathematical representation.

This thesis presents novel descriptive multidimensional Markovian textural models applied to computer aided diagnosis in the eld of X-ray mammography. These general mathematical models, applicable in wide areas of texture modeling outside X-ray mammography as well, provide ideal visual verication using synthesis of the corresponding measured data spaces, contrary to standard discriminative models. All achieved results in the thesis are extensively benchmarked. The thesis presents two methods for breast density classication in X-ray mammography. The methods were tested on the widely known MIAS database and the state-of-the art INbreast database, with competitive results. Several methods for completely automatic mammogram texture enhancement are presented. These methods are based on the descriptive textural models developed in the thesis which automatically adapt to the analyzed X-ray texture, thus being universal for any type of input without the need of further manual tuning of specic parameters. The methods' outputs highlight regions of interest, detected as textural abnormalities. The methods provide the possibility of enhancement tuned to specic types of mammogram tissue. Hence, the enhanced mammograms can help radiologists to decrease their false negative evaluation rate. It has been shown that the algorithms work well both for small ndings, such as microcalcications, and for bigger lesions. The pseudocolour method oers a unique way of mammogram feature fusion for visual evaluation and vastly enriches the the information content of the enhanced mammogram. The results were veried also by radiologist consultants. New contrast criterion was implemented which outperforms previously published contrast criteria. The focus of our study is mammogram texture and the following search for its optimal mathematical representation.