Automatic determination of regions of interest in chest images from magnetic resonance

• Grant number: 14/23307-7
• Support Opportunities: Scholarships in Brazil - Scientific Initiation
• Start date: December 01, 2014
• End date: November 30, 2015
• Field of knowledge: Engineering - Biomedical Engineering - Bioengineering
• Principal Investigator: Marcos de Sales Guerra Tsuzuki
• Grantee: Luciano Menasce Rosset
• Host Institution: Escola Politécnica (EP). Universidade de São Paulo (USP). São Paulo , SP, Brazil

Abstract

Multiple image processing algorithms, such as segmentation, registration and others, require the definition of a region of interest (ROI) in order to define the domain to be processed. The automatic determination of ROIs has been vastly investigated for the detection of vehicles' license plates. In most of the other cases ROIs are defined manually by the user. This manual definition is already convenient in many cases especially when the ROI needs to be defined only once. The candidate entered EPUSP in 2010 and was approved in the first semester's subjects. He was approved for the Molecular Sciences Course, with duration of four years, and started it in the second semester of the same year (2010). During the Molecular Sciences Courses the candidate enrolled for some of the Mechatronics Engineering course's subjects, moment in which the candidate came to know the current advisor. Currently the candidate is graduated in the Molecular Sciences Course and he is enrolled back in the Mechatronics Engineering course again. The definition of the ROI in medical images processing is a very important activity that determines wether the segmentation and registration algorithms, among others, will succeed or fail. Particularly, the lung is an organ with a large geometric variation, making its segmentation very complex. This scientific initiation project is part of a larger project, in which it is a necessary step. In this larger project it is desired to segment the lung's contours in multiple bidimensional images obtained through magnetic resonance in different planes. The automatic determination of the ROIs will allow the correct automatic determination of pulmonar contours. The algoritm to be studied in this scientific initiation will also aid in the segmentation of other adjacent organs such as: liver, heart and stomach. The ROI will be determined by a probabilistic heuristic (simulated annealing) with the definition of optimization functions. Some restrictions will be applied to build an anatomic atlas. Integral images will be used to accelerate the objective function computation.