CT computer aided diagnosis system

CT Computer-Aided Diagnosis System

PRESENTED BY

ABDALLA MOSTAFA ABDALLA

Scientific Research Group in Egypt http://www.egyptscience.net

7 March 2015 - Zewail University for Science and Technology

SRGE Members

Group founder and chair: Professor Aboul Ella Hassanien

Scientific Research Group in Egypt

Agenda Introduction

Problem Definition

Objective

Liver and Medical imaging

Pre-processing

Segmentation

Proposed Approach

Experiments and Results

Conclusion

Introduction

◦ Liver is an important organ in humanbody.

◦ It may have different colors (darkblue cyst, dark brown - cirrhosis,yellow - fatty, green – billarycirrhosis)

◦ It is common to use ComputedTomography (CT) in Computer aideddiagnosis systems (CAD)

Problem Statement

Difficulties associated with liver image segmentation

◦ Liver has different shapes.

◦ Similarities to other organs (muscles, flesh, kidney, spleen).

◦ Similarity between Vessels and cyst.

Objective

6

We aim to

• Enhance Region growing technique.

We have chosen

LiverCT

ImagesComputer

Manipulation

CADComputer-Aided Diagnosis System

Liver

8

Why Liver?

9

The statistics of liver diseases shows that• The ratio of virus C infection is 12.8 % in Egypt.• The ratio of virus C infection is almost 1.2% in

Europe.• 130 thousands people need liver transplantation

In Egypt.

Liver diseases

Cyst

Fatty Liver

Fibrosis

Billary Cirrhosis

Carcinoma

10

Different diseases may have different colors

According to• Liver bible• Pathology Atlas• Oncology ref.

So, Image can help in diagnosis

Biopsy

11

• It may puncture the lung.• It may fracture rib.• Liver bleeding.

• The worst of allthe sample might not represent the

lesion.

Biopsy has its limitations and risks

CT Image Slicing

12

Slicing technique

Liver sliced imageCT machine moves through the abdomen and records the details of liver tissues

Proposed approach

1• Preprocessing

2• Segmentation

13

The proposed approach has main two phases

Preprocessing

14

The main objective of image preprocessing is to improve the quality of the image being processed by:

• Removing noise.

• Emphasizing certain features.

• Isolating regions of interests.

Liver Segmentation

15

Liver segmentation depends on :

• The difficulty of the anatomy of liver.• Liver is surrounded by many organs, similar to its intensity as

spleen, stomach, and kidney.• The nature of liver tissues, and blood vessels.

.

Region Growing Segmentation

16

Based on the growth of a homogeneous regionaccording to certain features as intensity, coloror texture.

17

Now Let us go to the

Proposed Approach

Phases of Proposed Approach

18

Segmentation

Region growing

Preprocessing

Using morphological operations

Connecting Ribs

Morphological operationsMorphological Operations are :-

◦ Structure element.

◦ Dilation.

◦ Erosion.

◦ Opening.

◦ Closing.

19

Morphological operationsStructure element

has a shape of square, diamond and cross

20

Morphological operations

DilationThe basic effect of the operator on a binary image is to gradually enlarge the boundaries (thicking) of regions of foreground pixels.

21

Morphological operationsErosionThe basic effect of the operator on a binary image is to shrink(erode away ) the boundaries of regions of foreground pixels

22

Morphological operationsOpening

It is an erosion followed by a dilation.

It can open up a gap between objects connected by a thin bridge of pixels.

23

Morphological operationsClosing

is a dilation followed by erosion, it fills some gabs.

24

Connecting ribs

25

Using contrast stretching to emphasize the ribs boundaries. The ribs will be connected as follows:

Now the image is prepared for the next phase

Segmentation

26

•It is partitioning an image into homogeneous regions with respect to intensity, or texture.

•Image segmentation methods can be categorized as • Edge-based methods (discontinuity )• Region-based methods (similarity)

Liver Segmentation

27

So, there is a need to get

Separated Liver Regions of Interest

Proposed algorithm

28

Experiments and results

29

Cleaning image is a process of removing annotation and bed from the image

Experiments and results

30

Preparation phase

31

Similarity Index

Validation measure

Experiments and results

32

Difference betweensegmented and annotated image .

Experiments and results

33

Normal Region Growing vs Proposed approach

Experiments and results

34

Using the proposed method showed that: The accuracy result using similarity index measure is (SI=91.2% ). The method could segment images that was difficult to segmented before.

Conclusion

35

Testing proposed approach, with region growing showed that:

Normal Region growing has the result of 82% accuracy.Proposed approach has the result of 91.2% accuracy.

Future Work

36

The future work would be the change of the approach of classification to use Bio-Inspired methodology to :-

• Eliminate the liver separation computational cost.• Generalize the approach for other organs as spleen

and stomach.