Biomedical Image Processing & Detection of Cancer
An automatic tool for prediction and classification of cancerous/non-cancerous squamosal cells using image processing and artificial neural networks (ANN). The ANN program is much more flexible and user friendly. It also optimizes number of hidden nodes itself based on the prediction accuracy. The neural network used here is a two layer neural networks and it uses standard back propagation algorithm. The first layer is use for detection of nucleus, cytoplasm and background of the image.Whereas the second layer is used to classify images into cancerous or non-cancerous based on three cellular features: size of nucleus, size of cytoplasm and ratio of nucleus/cytoplasm sizes. Using image processing techniques we extracted 15 features for selected pixel (using a mouse event program written in DOT NET framework) of an image over its 3×3 neighbouring matrix (using program written in MATLAB and C). These parameters are input to first layer of ANN.
:: MORE INFORMATION