Parameter Optimization of Local Fuzzy Patterns Based on Fuzzy Contrast Measure for White Blood Cell Texture Feature Extraction

The parameter optimization of local fuzzy patterns based on the fuzzy contrast measure is proposed for extracting white blood cell texture. The proposed method obtains the optimal parameter values of the nucleus and cytoplasm region of white blood cell image and the best accuracy rate of white blood cell classification can therefore be achieved. To evaluate the performance of the proposed method, 100 microscopic white blood cell images and the supervised learning method are used for white blood cell classification. Results show that the average accuracy rate of white blood cell classification using local fuzzy pattern features with optimal parameter values of a nucleus and a cytoplasm region is 4% more accurate than with uniform parameter values and is 5-18% more accurate than other feature extraction methods. White blood cell feature extraction is part of the white blood cell classification in an automatic cancer diagnosis that is being developed. In addition, the proposed method can be used to obtain the optimal parameter of local fuzzy patterns for other types of datasets.