Breast Cancer Prognosis Using Image Processing
Abstract
Breast Cancer Prognosis Using Image Processing
Abstract
|
|
Breast Cancer Prognosis Using Image Processing Abstract |
|
Aims: This report utilises known prognostic indicators to develop a system to acquire data regarding the indicators and quantify them with the intention of using them to build a classification system using an intelligent machine to predict the prognosis of a given sample. Methods: Image processing methods were employed to segment cells from the background, information on the detected cells was used to quantify certain prognostic indicators. An Intelligent machine was then employed to classify and test the data. Results: Thirty test cases existed with four classes present within the test set, classes 3 and 4 proved quite accurate if the results were analysed using an out-by-one strategy (78%), with only one anomaly (1 class 4 case was classified as class 1). Classes one and two performed poorly, however only 5 cases existed between the two classes resulting in either the classier having not enough or nil data to build a classifier or no data to test the classifier. Conclusions: The segmentation algorithm didn't perform as well as expected on the real data (performed well using test images), this may have had an effect on the results, however the system produced results that were 39% accurate and 78% if the results were analysed using an off-by-one method. The data set needs to be larger with at least more examples of class 1 and class 2 cases to build a useful classifier. |
|
Home | Overview | Abstract | Main | Results | Future | References | Download | Terminology | Links | About The Author | Contact us Copyright © 2002 Nick Bryan All rights reserved. |