Pit Pattern Classification in Colonoscopy using Wavelets - WaveLab
Pit Pattern Classification in Colonoscopy using Wavelets - WaveLab
Pit Pattern Classification in Colonoscopy using Wavelets - WaveLab
Create successful ePaper yourself
Turn your PDF publications into a flip-book with our unique Google optimized e-Paper software.
1.3 Computer based pit pattern classification<br />
1.3 Computer based pit pattern classification<br />
The motivation beh<strong>in</strong>d computer based pit pattern classification is to assist the physician<br />
<strong>in</strong> analyz<strong>in</strong>g the colon images taken with a colonoscope just <strong>in</strong> time. Thus a classification<br />
can already be done dur<strong>in</strong>g the colonoscopy and therefore this makes a fast classification<br />
possible. But as already mentioned above, a f<strong>in</strong>al histological f<strong>in</strong>d<strong>in</strong>g is needed here too to<br />
confirm the classification made by the computer.<br />
The process of the computer based pit pattern classification can be divided <strong>in</strong>to the follow<strong>in</strong>g<br />
steps:<br />
1. First of all as many images as possible have to be acquired. These images serve for<br />
tra<strong>in</strong><strong>in</strong>g as well as for test<strong>in</strong>g a tra<strong>in</strong>ed classification algorithm.<br />
2. The images are analyzed for some specific features such as textural features, color<br />
features, frequency doma<strong>in</strong> features or any other type of features.<br />
3. A classification algorithm of choice is tra<strong>in</strong>ed with the features gathered <strong>in</strong> the last<br />
step.<br />
4. The classification algorithm is presented some unknown image to classify. The unknown<br />
image <strong>in</strong> this context is an image which has not been used dur<strong>in</strong>g the tra<strong>in</strong><strong>in</strong>g<br />
step.<br />
From these steps the very important question which features to extract from the images<br />
arises. S<strong>in</strong>ce there are many possibilities for image features, chapter 3 will give a short<br />
overview of some possible features for texture classification.<br />
However, as the title of this thesis already suggests, the features we <strong>in</strong>tend to use are solely<br />
based on the wavelet transform. Hence, before we start th<strong>in</strong>k<strong>in</strong>g about possible features<br />
to extract from endoscopic images, the next chapter tries to give a short <strong>in</strong>troduction to<br />
wavelets and the wavelet transform.<br />
5