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/

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

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.

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 :)

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

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.

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