Research On Classification Model For Complex Scenarios Of Biological Data

In recent years,Artificial Intelligence has made great strides in the field of medicine,especially in the field of image segmentation,where great progress has been made.In modern medicine,doctors often diagnose conditions based on different medical images and make corresponding treatment plans.Medical images are diverse,and each image has its own characteristics and areas of expertise.Such as computed tomography(CT),ultrasound imaging,magnetic resonance imaging(MRI),etc.With the help of machine learning algorithms,we use computer programs to process medical images of different modalities to help doctors improve diagnosis efficiency.In the treatment of cancer,radiotherapy is one of the most important means,and for the vast majority of cancers,radiotherapy is the last resort.In the process of radiotherapy,the determination of tumor target area is a key step,which requires multi-physician cooperation and discussion,which requires rich clinical experience.However,the tumor target areas drawn by different doctors will still be different.For this reason,we can use machine learning algorithms to automatically segment the edges of the lesions to assist doctors in the circle drawing work.Specifically,this paper mainly carries out the following research work:(1)Aiming at the problem of delineating the tumor target area in radiotherapy,we used the support vector machine(SVM)algorithm to achieve small sample classification,and used the KNN algorithm to achieve the idea of classifying all samples,and successfully achieved the kidney segmentation task in the human lower abdomen MR image..With the help of the dimensional processing of Kalman filter,the classification accuracy of the SVM algorithm and the accuracy of the kidney contour are greatly improved.Experiments demonstrate that our method achieves the highest DSC accuracy and obtains kidney contours that substantially overlap with manual segmentation.(2)A semi-supervised extreme learning machine(SSL-ELM)model is proposed,which uses the idea of semi-supervised classification and uses a membership function similar to the FCM algorithm,making full use of the correlation between unlabeled data improves the classification ability of the extreme learning machine model.It excels in the kidney segmentation task of MR images of the human lower abdomen,realizes the circle drawing of the kidney contour,and enhances the interpretability of the model.(3)The traditional linear weighted multi-core support vector machine uses the gradient descent method to determine the weights.This method requires multiple iterations to converge when it is close to the minimum value.We have developed an adaptive weighted multi-core support vector machine model(AW-SVM),which greatly reduces the number of iterations and improves the classification efficiency of the model.Experiments show that our algorithm achieves the highest DSC accuracy on the kidney segmentation problem for MR images of the human lower abdomen,which is comparable to the traditional excellent classification algorithms.