Machine Learning Predictive Model Of Prostate Puncture Based On Prostate MRI Image

Research background and purpose:Artificial intelligence technology is a technology that has developed rapidly in recent years.Through machine learning algorithms,it can analyze and process data that cannot be dealt with by conventional algorithms in the past.It can filter out the parts that are related to specific indicators from the messy sample data.The effect is stronger than the previous prediction model.This study uses MRI image data to establish a machine learning prediction model to predict the positive rate of prostate punctureMaterials and methods:Collected patients who underwent prostate puncture and MRI examinations in qilu hospital from 2010 to 2020,and their MRI examinations were performed within 1 week after admission.Collect the original data of the DICOM format of the MRI image,and select the lesion area on the T2 image with automatic,semi-automatic and manual correction.Calculate the radiomic characteristics of the lesion area through the pyradiomics plug-in.The dimension reduction algorithm is used to reduce and simplify the radiomics feature data,and filter out the data that can represent the difference.Finally,use neural network,random forest,support vector machine,logistic regression and other different machine learning algorithm models for training,and use the five-cross-validation algorithm to establish the prediction model.Finally,the comprehensive performance of the prediction model is evaluated through the area under the curve.Software enviroment:Use SPSS 23.0 and SAS9.4 to perform statistical analysis;use Matlab 2014b to program SVM,LRA and ANN code;use 3Dslicer4.13.0 to devide region of interest,to calculate radiomics features.Hardware environment:CPU intel corei5-9400f;memory DDR4 2666MHz 8GB*1;GPU Nvidia GTX 1650superResearch results:The prediction model trained by ANN algorithm can predict the potential positive rate of prostate puncture,followed by RF and SVM,but k-means algorithm can not be applied at all;In different prediction models,the importance of TPSA variables ranked first.Research conclusions:The machine learning model based on MRI images can accurately predict the positive rate of prostate puncture and can be used to predict prostate puncture.