Research And Application Of CTLM Assisted Diagnosis Of Breast Cancer Based On Machine Learning

Breast cancer is one of the highest incidence rate cancers in the world.It seriously threatens people’s health.Prevention is very important.Although breast cancer can be detected by X-ray,thermal imaging,ultrasound imaging and other methods at present,these detection methods may have uncomfortable effects on the body of patients and increase the workload of doctors.Computed tomographic laser mammography(CTLM)is a new method for breast cancer detection.At present,there is little research on the auxiliary diagnosis based on this method.This project comes from the product development project of a medical device company.It studies a kind of breast cancer auxiliary diagnosis system based on CTLM image,which has low cost,little harm to patients,can assist doctors to make judgments,improve the early diagnosis rate,and has very good application value.The specific research contents are as follows:1、The original image set is modeled by adaptive threshold partition,and sub blocks are divided to generate the basic data unit of this experiment.On the premise of using shape feature and texture feature separately,a combined feature set based on shape and texture is proposed,and the generated multiple feature sets are used as the subsequent model input variables.2、Four machine learning models(logistic,SVM,decision tree and BP)are selected to carry out comparative experiments on different features respectively.According to the experimental results,the advantages of combined features are determined.Through the basic evaluation of the results of four machine learning models,C4.5 decision tree has better model classification effect.3、Through the experiment of feature splitting,the advantages of the complete combined features are determined,and the integrated learning method is used to improve the individual machine learning method.The above four machine learning models are used as the base classifiers,and Ada Boost is used as the integrated learning method.By testing the experimental results of the model under different iterations,and comparing with other methods,the BP + boost model is determined Type A has the best classification effect in a certain algebra.4、With BP + boost model as the core classifier,a complete breast cancer auxiliary diagnosis system is designed.The system adopts the software architecture mode of layered architecture,and divides the system functions into several independent level layers,and shows the functions of the system software through the actual diagnosis of cases.The practice shows that the application effect of the developed system is good.