Parameter Estimation of Expectation Maximization Algorithm for Intensity Inhomogeneity Correction in Brain MR Images.

Date of Submission

December 2009

Date of Award

Winter 12-12-2010

Institute Name (Publisher)

Indian Statistical Institute

Document Type

Master's Dissertation

Degree Name

Master of Technology

Subject Name

Computer Science


Machine Intelligence Unit (MIU-Kolkata)


Maji, Pradipta (MIU-Kolkata; ISI)

Abstract (Summary of the Work)

Medical imaging is one of the powerful tool for gaining insight into the normal and pathological processes that affect health.Now a days the various imaging modalities, such as microscopy,computer tomography, ultrasound ,medical resonance imaging (MRI) and PET etc, are used in medical decision making processes and in surgical actions.Therefore high quality of accuracy is needed in taking the images .Clinical applications of a medical image require that image should be sufficiently clear and free from artifacts.That is why we need some preprocessing steps to remove the various image artifacts that comes due to imperfection in the image acquisition process.In this work, a class of preprocessing step ,will be addressed ,that deals with a spurious smoothly varying image intensity, which is apparent in the images obtained by different imaging modalities such as microscopy,CT,ultrasound and above all in the magnetic resonance imaging.This spurious variation of intensity is known as intensity inhomogeneity, intensity non-uniformity or bias field.Basically I will address the intensity inhomogeneity in MRI s as the impact of this image technique in neurological applications is impressive, due to less side effects and flexibility in joining high-quality anatomical images with functional information.1.1 Road-map of the reportIn chapter 2, basic principle of MR Imaging is discussed .Here I have briefly covered the advantages and disadvantages of MRI s .Because of its usefulness for the soft tissues, it is widely used for taking the image of human brain . In chapter 3, intensity inhomogeneity ,its causes and the basic models of intensity inhomogene7 ity in Medical Resonance Imaging are discussed .Also the various approaches of intensity inhomogeneity correction are discussed. The summary of works done on bias field correction till today, is briefly covered in this chapter .In chapter 4, a correction strategy is discussed ,which is based on expectation maximization and log likelihood estimation .Where I have given the complete mathematical setup of this approach and have highlighted the various issues such as the tuning of parameters of the algorithm, related to the algorithm.In chapter 5, C-Means Clustering algorithm, that is used in the estimation mean and variance of tissue classes, discussed .Both the Hard C-means and Fuzzy C-Means algorithms are discussed in detail .As these are used in inhomogeneity correction work for estimating mean and variance of tissue classes.Chapter 6 consists of my core work ,where I have tried to tune the parameters of this correction approach using the bench marked data sets, I have put the results of my various experiments in which I used various strategies for estimating the parameters of the image and the bias field to optimize the performance of this algorithm.Estimated parameters are used to correct the real life MR images.Some examples of real MRI are also given this chapter.Chapter 8 contains the Summary and Future scope of this work.And finally the appendix contains some of the c-code used for the experiment purposes.


ProQuest Collection ID:

Control Number


Creative Commons License

Creative Commons Attribution 4.0 International License
This work is licensed under a Creative Commons Attribution 4.0 International License.


This document is currently not available here.