Survey and Analysis of Mathematical Methods in Texture Classification with Application to Ultrasound

Student: Florian Schulze
Supervisor: Christian Wachinger
Director: Nassir Navab

Ultrasound images are generally considered to be gradient-like images, meaning that mainly the boundary between the organs is visualized. Frequently, not much intention is paid to the interior region of the organs, which is dominated by speckle noise. A common strategy, like in project 1, is to reduce the speckle noise by filtering operations. In this project we would like to investigate the possibility for learning the texture of organs. This is motivated by the fact that speckle is not any arbitrary kind of noise, but is the result of micro-inhomogeneities in the tissue. Since there is a physical reason behind the appearance of speckle, the extraction of a characteristic texture for each organ seems feasible. The student will first have to read into different methods of texture analysis and synthesis, which are closely related to machine learning techniques. In a second step, a few methods will be tested on real ultrasound images which we will acquire with our ultrasound system at the hospital rechts der Isar. Requirements: Knowledge in programming, optionally experience in Matlab Kwonledge in image processing Interested in applying machine learning and statistical methods to real data Interested in ultrasound imaging

Courtesy of Yanxi Liu (CMU)