Mathematical models for analyzing and interpreting microwave radiometry data in medical diagnosis

The present paper is devoted to the improvement and application of mathematical models for the dynamic description of the patient state in the diagnosis of breast cancer and venous diseases based on microwave radiometry data. I present and describe in detail a modified approach for constructing interpretable features in thermometric data. A model evalutation is performed by constructing classification algorithms in the following feature spaces: temperature values, thermometric features, 2nd, 3rd and 4th degree polynomial features. Best algorithms have sensitivity value of 0.892 and specificity value of 0.813 in the mammary glands dataset and sensitivity value of 0.961 and specificity value of 0.925 in the lower extremities dataset. The algorithms built also provide an explanation of result in terms which are understandable for clinicians. The most important features in thermometric data are presented, as well as an example of explanation building.