Medical image analysis

Medical Image Analysis

Hossam Mahmoud Moftah and Aboul Ella Hassanien

Cairo University,

Dept. of Information Technology, Faculty of Computers and information

Scientific Research Group in Egypt

http://www.egyptscience.net

2

Agenda • Introduction

• Objectives

• Scope of work and proposed solutions

• Medical Imaging

• Medical Image Segmentation

• Ant Colony Optimization

• Proposed Approaches

• Adaptive K-Means Clustering Algorithm for MR Breast Image Segmentation

• 3D Brain Tumor Segmentation Scheme using K-mean Clustering and Connected Component Labeling Algorithms

• Volume Identification and Estimation of MRI Brain Tumor

• MRI Breast cancer diagnosis hybrid approach using adaptive Ant-based segmentation and Multilayer Perceptron NN classifier

• Conclusions and Future Work

3

Introduction

• Bio-medical Image analysis and processing has great significance in the field of medicine, especially in Non-invasive treatment and clinical study.

• There are many problems in medical image analysis and interpretation involve the need for a computer aided system to understand the images and image structure and know what it means.

• The accurate interpretation and analysis of medical images often become boring and time consuming, because there is much detail in such images.

4

Objectives

• Developing computational methods and algorithms to analyze and quantify biomedical data.

• Applying information analysis and visualization to biomedical research problems.

• Developing methods and applications to give our collaborators the ability to analyze biomedical data to help in the discovery of biomedical knowledge and the diagnosis of diseases.

5

Scope of work and proposed solutions

• In this research complete solutions are proposed for medical image analysis.

• Different approaches and systems that combines the advantages of intelligent techniques are presented such as:

• Ant-based clustering, k-means clustering and neural

network classifiers, in conjunction with statistical-based feature extraction.

• Robust medical imaging systems to analyze and interpret 2D and 3D medical images including 3D brain tumor, MRI breast images.

6

Medical Imaging

• Image intensities can be:

• Radiation absorption in X-ray imaging

• Acoustic pressure in ultrasound

• Radio frequency (RF) signal amplitude in MRI

• Dimensionality: Refers to whether a segmentation method operates in a 2-D image domain or a 3-D image domain.

• Generally, 2-D methods are applied to 2-D images, and 3-D methods are applied to 3-D images.

7

Medical Imaging

• In some cases, however, 2-D methods are applied sequentially to the slices of a 3-D image.

• This may arise because of practical reasons:

• Ease of implementation

• Lower computational complexity

• Reduced memory requirements

• Major Modalities:

• Projection X-ray (Radiography)

• Computed Tomography (CT)

• Magnetic Resonance Imaging (MRI)

• Ultrasound

8

MRI vs. CT Scan

• CT scans are a specialized type of x-ray

• MRI uses a magnetic field with radio frequencies introduced into it.

9

MRI vs. CT Scan

MRI CT

MRI Scan show tendons and ligaments very well

CT is not the best choice for that

MRI is not the best choice for that

Bleeding in the brain, especially from injury

Tumor in the brain is better seen on MRI CT is not the best choice for that

MRI is not the best choice for that

bone structures - the inner ears - can easily detect tumors within the auditory canals

MRI is not the best choice for that

CT shows organ tear and organ injury quickly and efficiently for the damaged organs or torn in accident

MRI is not the best choice for that

Broken bones and vertebral bodies of the spine

Injury to the spinal cord itself MRI is not the best choice for that

10

MRI vs. CT Scan

11

Image Segmentation

• Image segmentation is the task of splitting a digital image into one or more regions of interest.

• Image Segmentation Techniques:

• Region-based segmentation

• Data clustering

• Edge-base segmentation

12

Image Segmentation

• Region-based segmentation

• Seeded Region Growing

• Region Splitting and Merging

• Data clustering

• Hierarchical

• Hierarchical divisive algorithm

• Partitional

• K-means Clustering Algorithm

• Edge-base segmentation

• Watershed Segmentation Algorithm

13

Ant Colony Optimization

• Ant Colony Optimization is an efficient method to finding optimal solutions to a graph

• Using three algorithms based on choosing a city, updating pheromone trails and pheromone trail decay, we can determine an optimal solution to a graph

• Ant Colony Optimization has been used to figure out solutions to real world problems, such as truck routing

14

Proposed Approaches: 2D Segmentation

Adaptive K-Means Clustering Algorithm for MR Breast Image Segmentation

Hossam M.Moftah, Ahmed Taher Azar, Eiman. T. Al-Shammari, Neveen. I.Ghali, Aboul Ella Hassanien, and Mahmoud Shoman, "Adaptive K-Means Clustering Algorithm for MR Breast Image Segmentation", Neural Computing and Applications Journal, DOI 10.1007/s00521-013-1437-4 (Springer), 2013. (Impact factor =1.6).

15

Adaptive K-Means Clustering Algorithm for MR Breast Image Segmentation

• The most popular method for clustering is k-means clustering.

• This article presents a new approach intended to provide more reliable magnetic resonance (MR) breast image segmentation that is based on adaptation to identify target objects through an optimization methodology that maintains the optimum result during iterations.

• The proposed approach improves and enhances the effectiveness and efficiency of the traditional k-means clustering algorithm.

16

K-means Algorithm

17

Block diagram of the proposed approach

18

Adaptive k-mean segmentation approach

19

Experimental results and comparative analysis

20

Experimental results and comparative analysis

21

Proposed Approaches: 3D Segmentation

3D Brain Tumor Segmentation Scheme using K-mean Clustering and Connected Component Labeling Algorithms

Hossam M. Moftah, Aboul Ella Hassanien, Mohamoud Shoman: 3D brain tumor segmentation scheme using K-mean clustering and connected component labeling algorithms. IEEE International conf, on Intelligent system and design applications ISDA 2010: 320-324.

22

3D Brain Tumor Segmentation Scheme using K-mean Clustering and Connected Component Labeling Algorithms

• In this article, an image segmentation scheme is proposed to segment 3D brain tumor from MRI images through the clustering process.

• The clustering is achieved using K-means algorithm in conjunction with the connected component labeling algorithm to link the similar clustered objects in all 2D slices and then obtain 3D segmented tissue using the patch object rendering process.

23

Block diagram of the proposed approach

24

Clustering and relabeling

25

3D Tumor segmentation algorithm

26

Experimental results

27

Proposed Approaches: 3D Analysis

Volume Identification and Estimation of MRI Brain Tumor

Hossam M. Moftah, Neveen I. Ghali, Aboul Ella

Hassanien, Mahmoud A. Ismail: Volume identification and estimation of MRI brain tumor. IEEE Hybrid Intelligent system (HIS 2012): pp. 120-124.

28

Volume Identification and Estimation of MRI Brain Tumor

• This article deals with two dimensional magnetic resonance imaging (MRI) sequence of brain slices which include many objects to identify and estimate the volume of the brain tumors.

29

Volume measurement: an illustrated example

30

Volume estimation

31

Experimental results

32

Proposed Approaches: 3D Analysis

MRI Breast cancer diagnosis hybrid approach using adaptive Ant-based segmentation and Multilayer Perceptron NN classifier

Hossam M. Moftah, Ahmed Taher Azar, Aboul Ella Hassanien and Mahmoud Shoman, MRI Breast cancer diagnosis approach using adaptive Ant-based segmentation and Multilayer Perceptron NN classifier. Applied Soft Computing Journal (Elsevier), 2013. (Accepted Impact factor = 2.5).

33

MRI Breast cancer diagnosis hybrid approach using adaptive Ant-based segmentation and Multilayer Perceptron NN classifier

• This article introduces a hybrid approach that combines the advantages of fuzzy sets, ant-based clustering and Multilayer Perceptron Neural Network (NN) classifier, in conjunction with statistical-based feature extraction technique.

• An application of breast cancer MRI imaging has been chosen and hybridization system has been applied to see their ability and accuracy to classify the breast cancer images into two outcomes: Benign or Malignant

34

Block diagram of the proposed approach

35

Adaptive Ant-based segmentation algorithm

36

Experimental results and comparative analysis

37

Experimental results and comparative analysis

38

Experimental results and comparative analysis

39

Experimental results and comparative analysis

40

Conclusions

• Bio-medical image analysis solutions and systems are presented in this thesis.

• An efficient 2D and 3D segmentation algorithms for medical images are presented to solve medical image segmentation problems.

• A new approach is presented intended to provide more reliable MR breast image segmentation.

• An image segmentation scheme is presented to segment 3D brain tumor from MRI images.

• 3D brain tumor identification and volume measurement algorithm is presented for brain MRI.

• A hybrid approach is presented that combines the advantages of fuzzy sets, ant-based clustering and Multilayer Perceptron Neural Network (NN) classifier, in conjunction with statistical-based feature extraction technique.

41

Future Work

• Introducing techniques and algorithms to solve the traditional medical problems in my country Egypt such as the early detection of liver fibrosis stages and other medical problems.

• Overcoming limitations inherent in conventional computer-aided diagnosis, and try to apply intelligent methods and algorithms such as rough sets, near sets to present an effective method of dealing with uncertainties.

• Introducing and applying optimization techniques to solve medical problems such as multi-objective optimization or pareto optimization.

• Introducing new versions of optimization algorithms such as Discrete Invasive Weed Optimization (IWO) algorithm inspired from weed colonization to solve medical imaging problems.

• Dealing with animal medical images to solve medical animal problems such as mammary gland tumors in cats.

• Searching for different new machine learning techniques to optimize the classification step.

42

Thanks

http://www.egyptscience.net