L.T.D

kd tree

2010-04-16T04:25:00.000-07:00

SA,
how are u girls? hope you are fine :)

our objective

Is to retrieve most similar cases(images) to the one queried by doctor to know its diagnosis
Having a collection of feature vectors stored in a database in a d-dimensional space, image retrieval is done using k-nearest neighbors algorithm.
The most straight forward way is to sequentially compare all points using the given distance measure.
This linear search takes O(n) time, which will be so slow for large databases

K-d tree

kd-tree or k-dimensional tree is a space partitioning data structure for organizing points in a k-dimensional space.
kd-trees are efficient methods of finding nearest neighbors in high dimensional spaces

Resources
To read more..
The wikipedia link: http://en.wikipedia.org/wiki/Kd-tree
you can find kd tree implementation here: http://www.cs.wlu.edu/~levy/software/kd/

Phase 1

2010-03-18T03:44:00.000-07:00

SA,
Thanks to Allah phase 1 has been accomplished, and we (Fatma and Ahd) preferred to document this phase, to organize our work and to help us in the final documentation isA.

The documentation is available here

http://www.fileden.com/files/2007/6/9/1159927/Graduation%20Project/phase%201.pdf

Second seminar

2010-03-18T02:38:00.000-07:00

Our second seminar was on 16/3/2010
Thanks to Allah it was a good one :)

We first started with a quick review on our motivation, objective and system overview, then we talked about the accomplished phase which consists of 2 main steps (Lung segmentation and tumor segmentation).
Finally we mentioned the remaining two phases as our future work.

That's the presentation we used in this seminar.

L.T.D second seminar

View more presentations from FatmaSamy.

Flood fill algorithm - Demo

2010-02-16T10:09:00.000-08:00

SA,
Flood fill algorithm is an algorithm used to fill an area of certain color with another color.
We use it in our application after extracting the lung borders to refill the lung area with the original intensity of the source image.
Here's a little demo for this algorithm, the image is fixed, and don't spend much time asking your self what this image could represent :D
it's th image resulted from sobel edge detection of lung CT image, anyway :D
all you have to do is to choose a color and click the area you want to fill, and get the image colored ;)
It's a direct implementation of the third pseudo code that could be found here
http://www.codecodex.com/wiki/Implementing_the_flood_fill_algorithm

My implementation is downloadable from here
any comments or questions are appreciated :)
Thanks :)

Insight Segmentation and Registration Toolkit (ITK)

2010-02-15T18:41:00.000-08:00

SA,
ITK is an open source system that provide developers lots of ready to use tools for image processing especially medical images segmentation and registration.
It's really useful I advise whoever going to deal with medical images to use this toolkit it's helpful (Y).

But since it's implemented in C++ and we use C# in our project, I used a wrapper called ManagedITK which covers the majority of ITK functions.

For more info please refer to:
---------------------------------------
http://www.itk.org/ ---> The ITK project

http://code.google.com/p/manageditk/ ---> The ManagedITK

Redefine our goal

2010-02-15T18:26:00.000-08:00

SA,
It has been while I know :)
But here we're again ;)

I don't know if the title is expressive enough !! But it's ok :D

I just felt that we need to redefine the goal of the project and the output we aim to gain. May be this is because we read more and asked more, so actually I feel that we keep discovering what we really have to do. But I promise we won't keep discovering till the Final seminar isA (A) :D

Well let me sum up what we aim to :
----------------------------------------------
- CBIR based CAD system for lung tumors.
- The system should be totally automated so that the user (physician) will not have to do much work in order to make the system able to identify the tumors.
- The automation is achieved through:
   - Automatic lung segmentation
   - Automatic identification of tumor candidates.
   - Automatic feature extraction.
   - Evaluation of the tumor characteristics which gives an initial diagnosis of the tumor (Malignant or Benign).
   - By retrieving similar tumors we can confirm our initial diagnosis.
   - Finally we show the physician the previous cases and probability that this tumor can be malignant or benign.

CAD system for early diagnosis of the Lung tumors in CT

2009-12-10T14:08:00.000-08:00

SA :)
This is a summary for a paper that proposes an automatic computer-aided diagnosis (CAD) system
for early detection of lung cancer by analyzing chest 3D computed tomography (CT) images.
This is a very useful paper, it discusses the steps followed by the system and states the algorithms used briefly.

For the summary : Click here
For the original : Go here and then download the full issue, the paper's name is:
"Computer Aided Diagnosis System for Early Detection of Lung Cancer Using Chest Computer Tomography Images"

It's a good site by the way with many papers available for free :D

CBIR in medicine General survey

2009-12-10T13:50:00.000-08:00

SA :)

This documentation is a general survey collected from recent papers in CBIR in medicine, this document is collected by me (Fatma) and Reham

Enjoy it ;)

View the document

The project idea

2009-12-06T08:31:00.000-08:00

Briefly, we aim to develop a system that can help physicians in diagnosing Lung tumors from CTs, depending on previous cases that are stored in the database.

This is the presentation of our seminar that give a good overview of the project

enjoy :)

First seminar presentation

View more presentations from FatmaSamy.

Active contour model (SNAKES) 1

2009-12-01T06:10:00.000-08:00

السلام عليكم و رحمة الله و بركاته

Before i begin talking about SNAKES, first i have to answer some questions.

What is image segmentation?
- It's subdividing or partitioning an image into it's constituent regions.

Why?
- To extract features of interest in images by looking for mathematical entities which describe the shapes of objects appearing in images.

So, why do we need image segmentation in our project?
- we want to extract the lung area from a CT (Computed Tomography) image which is our interest, for computing and detecting the abnormal shadow areas that are suspicious to be tumors.

What we have to do?
-we have to find a method that looks for any shape in the image that is smooth and forms a closed contour.
-From here we begin our research in "SNAKES".

Active contour model:

Introduction:

The active contour model algorithm, first introduced by Kass et al., deforms a contour to lock onto features of interest within an image.
Usually the features are lines, edges, and/or object boundaries.
Kass et al. named their algorithm, “Snakes” because the deformable contours resemble snakes as they move

Framework for Snakes

The active contour system provides the ability to interactively guide the contour detection by placing initial points
A higher level process or a user initializes any curve close to the object boundary.
The snake then starts deforming and moving towards the desired object boundary.
In the end it completely “shrink-wraps” around the object.

To be continued isA....

CBIR ??!!

2009-11-28T13:08:00.000-08:00

CBIR .. what does this stand for ?!
CBIR stands for Content Based Image Retrieval.
and in brief CBIR is a technique used in image retrieval from large image databases, instead of relying on on metadata such as captions or keywords in retrieving images which is expensive while saving images and inaccurate in retrieval.
"Content" may be ambiguous, to be clear content of an image may be colors in it, textures, shapes or any other information that can be extracted from an image.

The following figure shows a general structure of CBIR systems

CBIR systems always consists of 2 phases the offline phase and the offline phase

where “offline” indexing phase is displayed in the bottom part. Usually the visual contents of the images in the database are extracted and described by multidimensional feature vectors. A feature database is formed using the feature vectors of the images in the database. The “online” content-based retrieval is displayed in the upper part. Users provide the retrieval system with query example images, which are used to retrieve images.
The system then extracts the feature vectors from the example images. The similarities/distances between the feature vectors of the query example and those of the images in the database are calculated. The online and offline phase interact with a collection of multimedia items (images, videos, etc.) from a multimedia database. The query provided by the user can be an example image, region, sketch, humming, or text .

Medicine is one of the most potential application area.

With medical imaging techniques such as X-Ray, computer tomography, magnetic resonance imaging, and ultrasound, the amount of digital images that are produced in hospitals is increasing incredibly fast. Thus the need for systems that can provide efficient retrieval of images of particular interest is becoming very high.

Since for very large database personally describing and annotating every image with text indices is time-consuming and impractical (TBIR - Text Based Image Retrieval or also called Metadata approach), CBIR seems to be a good approach.
Researchers are still working in this field, it's a new born field.
and thus we'll participate in researches and applications in this field isA :)

Resources:
1- Ivica Dimitrovski, Dejan Gorgevik, Suzana Loskovska, WEB-BASED MEDICAL IMAGE RETRIEVAL SYSTEM

2- http://en.wikipedia.org/wiki/Content-based_image_retrieval

Who are we ?!!!

2009-11-14T06:42:00.000-08:00

السلام عليكم ورحمة الله وبركاته

Who are we ??!

well we're three girls at the final grade in the Faculty of Computer and Information Systems (FCIS) Ain Shams University (ASU)

Fatma Mohammad Samy.

Reham Kamal.
Ahd Abdelrazek.

We made this blog for our graduation project, all about our GP will be available here isA, the idea, papers, websites and some of our work too isA ;)

Hope this will be useful for any who's searching in the same topic :)