BREAST CANCER CLASSIFICATION USING ANN Ross Nordstrom Intro
BREAST CANCER CLASSIFICATION USING ANN Ross Nordstrom Intro to Artificial Neural Networks Yu Hen Hu December 14, 2010
Overview Learning Methods Training Methods • K-Nearest Neighbor • Leave One Out • Maximum Likelihood (using negative log) • Random grouping over ~5%-95% of data set used for training • Used 10+ different sets for each ratio
![About Existing Approaches ~Leading Global Cause of Death in Women[1] ~1. 302 M cases](http://slidetodoc.com/presentation_image_h2/5c0e80c27be89719434ddb128de57c77/image-3.jpg "About Existing Approaches ~Leading Global Cause of Death in Women[1] ~1. 302 M cases")
About Existing Approaches ~Leading Global Cause of Death in Women[1] ~1. 302 M cases Basic ~ 97. 5% [2] - 1 separating plane in 3 D space of worst area/smoothness ~460 K deaths ~30% in developed ~45% in 3 rd world ~ Treatment based on the stage of the cancer ~Not easily detectable. Early signs may be hidden or attributed to hormones ~Mammograms detect 85 -90% * Has correctly diagnosed 176 consecutive new patients Diagnosis ~95. 6%[4] - Parallel hyper planes ~93. 7%[4] - 1 nearest neighbor Prognosis ~ 86. 3%[2] - Multisurface Method -Tree in 4 D space
Maximum Likelihood Leave-One-Out Training
Maximum Likelihood Randomly Distributed Training Set (2%-98%) Peak : 100% Best avg: 91% Avg: 89% Peak : 100% Best avg: 91% Avg: 86% Peak : 100% Best avg: 80% Avg: 64%
K-Nearest Neighbor Leave-One-Out Training 1 -15 Neighbors Mean: Median: Std: Basic Diag 81% 86% 80% 87% 2. 25 3. 40 Prog 68% 69% 4. 79
K-Nearest Neighbor Randomly Distributed Training Set (2%-98%) 1 -10 Neighbors
Conclusions ~Existing results are reachable ~Prognosis is the most difficult ~Better accuracy on groups than anecdotes ~Plane Separation > Max Likelihood > K-Nearest
![References [1] Garcia M, Jemal A, Ward EM, Center MM, Hao Y, Siegel RL,](http://slidetodoc.com/presentation_image_h2/5c0e80c27be89719434ddb128de57c77/image-10.jpg "References [1] Garcia M, Jemal A, Ward EM, Center MM, Hao Y, Siegel RL,")
References [1] Garcia M, Jemal A, Ward EM, Center MM, Hao Y, Siegel RL, Thun MJ. Global Cancer Facts & Figures 2007. Atlanta, GA: American Cancer Society, 2007. [2] Wolberg, Dr. William H. "Breast Cancer Wisconsin (Diagnostic) Data Set. " UCI Machine Learning Repository(1995): n. pag. Web. 30 Nov 2010. [3] Zielinski, Jerzy, Nidhal Bouaynaya, and Dan Schonfeld. "Two-dimensional ARMA modeling for breast cancer detection and classification. ” Engineering Village (2010): n. pag. Web. 30 Nov 2010. [4] Zwitter, Matjaz, and Milan Soklic. "Breast Cancer Data Set. "UCI Machine Learning Repository (1988): n. pag. Web. 20 Nov 2010.
- Slides: 10