Machine Learning Based Prostate Pathological Biopsy Image Recognize
|School||South China University of Technology|
|Course||Signal and Information Processing|
|Keywords||Prostate Machine Learning Deep Neural Network Spatial Pyramid|
As one of the common cancer among the elderly men, prostate cancer is main reasonwhich harm the health of the elderly men. Early detection and early treatment are helpful topatient. Automatic analysis algorithms of prostate pathological biopsy image has become thefocus of Domestic and foreign scholars. The development of Prostate computer aideddiagnosis system is very meaningful, which can be automatic classification of prostatepathological biopsy image and Auxiliary medical diagnosis.In order to improve efficiency and prevent prostate pathological image from artificiallymarking by mistake,This paper presents a method based on Deep Learning for prostatepathological image judge.This algorithm can be used to determine whether the pathologyimage comes from the prostate tissue. In this algorithm,we can sample3030image block inprostate pathological images randomly as input data.At first, we preprocess image block withZCA Whitening to get rid of the redundant data. Then Deep learning algorithms are used tosample image block with unsupervised learning. Features got at last step are as the input forsoftmax regression classifier, classifying image block of prostate pathological images. Finally,we classify the whole prostate pathological image based on ROC curves.In prostate cancer pathological image recognition, prostate cancer pathological image areobvious changes relatively to the normal prostate pathological image in global structuralinformation.1.stroma structure disorder and stroma direction is irregular；2. Gland lumen areabecome smaller, even disappear；3. The nuclei are scattered distribution. This paper presents amethod to describe spatial information of prostate pathological image based on SpatialPyramid Matching of BoW model. Finally, we recognize prostate cancer pathological imageby Support Vector Machine.The experimental results show that prostate pathological image analysis methodproposed in this paper is feasible and good robustness.