Extraction Of Palm Features From Female With Breast Cancer And Its Susceptibility Study

Early diagnosis of breast cancer is the key to improve the cure rate of breast cancer.Biological studies have shown that there is a correlation between the palm characteristics and the incidence of breast cancer.If we can extract the susceptible characteristics of breast cancer from the palm,it will provide a non-invasive and simple method for the detection of breast cancer susceptibility,and timely identify the early susceptibility,so as to improve the survival rate of breast cancer patients.Most of the existing studies on the correlation between palm features and breast cancer are manual feature extraction,which is time consuming and subjective.This paper studies the automatic extraction of palm characteristics and its correlation with breast cancer,It extracts the palm features including: a-b ridge line number(The number of ridge lines crossed from the center to the trigonometric point in a fingerprint),finger length ratio,palmprint mainline type and ATD angle(Refers to the Angle formed by the three points of the finger root and the palm root).It studies the relationship between these characteristics and breast cancer,screening the finger length ratio of 2D: 4D(Index finger’s length divided by ring finger’s length),palmprint mainline type,the ATD angle and the number of a-b ridge line as the sensitive characteristics.Subsequently,it builds a classifier to complete the breast cancer susceptibility prediction.The main contributions of the paper are as follows:(1)Data set establishment: A palm image acquisition system is designed through MFC,including the image acquisition interface,data query interface and image deletion interface.Then,I went to Yunnan tumor hospital to collect the palm images of breast cancer patients through fingerprint collector and digital camera.Then I collected the palm images of normal females as the control group,and established two databases of the sick group and the control group.(2)For feature extraction,this paper proposes a main line extraction of palmprint based on multi-direction filtering and neighborhood de-noising.Firstly,the palm image is preprocessed to uniform size and extract ROI(region of interest)of gray palmprint image.Subsequently,the image is denoised by a median filtering and performed a four-direction detector for ROI image filtering.The filtering results is applied bottom cap operation and a crude palmprint mainline is obtained.The finally mainline is achieve by binarizing and neighborhood denoising.Experiment results show that the proposed method can extract complete mainline of palmprint with no breakpoint and no noise.(3)After the characteristics of ATD angle,a-b ridge number,palmprint line mainline type and finger length ratio are extracted,the statistical analysis of these characteristics with the correlation to the breast cancer is carried out.The characteristics with significant sensitivity between the data of the sick group and the control group are screened out,including the finger length ratio of 2D:4D,a-b ridge number,the mainline type of palmprint and the ATD angle.(4)An adaboost classification algorithm based on the variant logistic regression step length is proposed.Firstly,the data to be classified is assigned a weight,and then it is classified by a logic regression algorithm with gradually reduced step size.Then the weights of the updated data and the classifier are iteratively regressed for classification.Finally,the regression classifier is weighted and combined to get the final classifier output.This classifier is used to susceptibility classify between breast cancer patients and normal people with the accuracy of more than 80%.