一种心脏MRI图像半监督学习语义分割算法OA

Due to the scarcity of medical image data and the difficulty to obtain labeled data,a semantic segmentation method for semi-supervised learning based on contrastive learning and hard sample recall loss is proposed,as well as a non-intrusive target regions cutout(NTRC)data enhancement algorithm.By designing the hard sample recall loss function,correcting the global error of binary cross entro-py loss function(BCE)and imbalance contrastive loss function(ICLF),the model is guided to find the optimal solution better.Mean-while,the NTRC data enhancement algorithm is used to enhance the data so that the area to be segmented is not invaded and damaged.This algorithm is used for semi-supervised learning semantic segmentation of cardiac MRI images.The results show that the algorithm solves the problem of the imbalance of positive and negative samples and the low accuracy of hard sample segmentation,and generally improves the performance of the semi-supervised semantic segmentation algorithm of cardiac MRI images.