Early diagnose breast cancer with PCA-LDA based FER and neuro-fuzzy classification system

In recent years, breast cancer is recognized as critical disease that needs to be detected at early stage. Breast cancer can be detected either by mammography or biopsy technique. Biopsy is cost effective procedure for breast cancer detection. Fine needle aspiration (FNA) digital image is pre-processed and extracted features from it used for classification. Based on the parameters of feature attributes, they are classified into two classes: malignant (cancerous) and benign (not cancerous) tumor cell. Manual analysis and classification of this image is very difficult and challenging task. The need of automation for detecting and classifying these tumor cells in cost effective and accurate manner provokes many researchers in this field. With this aim we proposed PCA-LDA based feature extraction and reduction (FER) technique that reduce the original feature space to large extent and performing training over this reduced set that give excellent accuracy of 98.6%. For classification we use ANNFIS classifier that uses the neural-network concept with some fuzzy rule logic. We perform comparative performance analysis study amongst our proposed work over two other classifiers i.e. support vector machine (SVM) and multi-layer perceptron (MLP). The experimental result shows that proposed framework outperform over SVM and MLP with an accuracy of 98.6%.

Automated summary

Early detection of breast cancer is crucial, with biopsy being a cost-effective method for detection. A proposed PCA-LDA based feature extraction and reduction technique achieved 98.6% accuracy in classifying tumor cells. Performance analysis showed that the proposed framework outperformed other classifiers like SVM and MLP.