Fuzzy-neural machine with image feature extraction for colorectal cancer diagnosis

An automated algorithmic approach, based on quantitative measurements, is a valuable tool to a Pathologist for fast verification of colon cancer image abnormalities for effective treatment. In this paper a novel method which automatically locates differences in colon cell images and classifies the colon cells into normal and malignant cells is presented. The system fuzzifies image feature descriptors and incorporates a clustering paradigm with neural network to classify images. The novelty of the algorithm is that it is independent of the feature extraction procedure adopted and overcomes the sharpness of class characteristics associated with other classifiers. It incorporates feature analysis and selection and differs markedly from other approaches which either ignore them or perform them as separate tasks prior to classification. The innovative method has been evaluated using 116 cancerous and 88 normal colon cell images and resulted in a very high classification rate of 96.435%. The percentage error rate of 2.6% is primarily due to preprocessing anomalies. The proposed system was evaluated using 116 cancer and 88 normal colon cell images and shown to be more efficient, simple to implement and yields better accuracy than other methods.