Medical Image Analysis Using Soft Computing Feature Selection and Classification of Skin Cancer

Adaptive resonance theory ARTI neural network (unsupervised learning) is being implemented for the classification of feature values with the objective of carrying out a comparative study with the performance of other technique including multi-layer feed forward network transform (supervised learning). A neural network is applied for the detection/identification of melanoma skin cancer. The images were collected from melanoma database is first processed and some distinguishable features were extracted by a MATLAB GUI. Two classes of the images were used: (a) benign, (b) malignant. A collection of interpreted binary input data, which is digitized from analog data after normalization is, then clustered adaptively with neural network ART-1. The goal of the project is to classify the feature dataset using supervised and unsupervised neural networks. To collect the feature set 1 implemented MATLAB image processing toolbox to create my own MATLAB functions. I created my own ARTI codes in MATLAB 701 (AKTI) to analysis which features are giving better efficiency using different vigilance parameter. Along with these supervised and unsupervised classification and pattern recognition, I am also trying to classify my own data-set of extracted features on Weka machine learning benchmarking software which is nothing but implements machine learning algorithms in JAVA.