Optimal Complexity Structures of Ultrasound Image Models in Diagnostic Decision Making System

The work considers medical images as objects which classification can be solved by set classification methods. Based on the algorithms developed by the authors to find the optimal structures of models for representing classification objects, it is proposed to describe ultrasonic images as nonlinear models of linear scanning and sliding window models. The obtained model parameters were used as texture features of the region of interest. In the result feature ensemble was used as input of classifier build on the Random Forest method. The quality high values of norm-pathology classification were obtained on the test sample of patients with diffuse liver diseases. Data for research were provided by SI “Institute of Nuclear Medicine and Radiation Diagnostics of the National Academy of Medical Sciences of Ukraine”.