摘要
脑部肿瘤MRI图像作为一种无介入性的诊断工具,包含极其丰富的病理、生理和解剖信息,为临床诊断工作提供了巨大的支持,在医学研究和疾病治疗中发挥重要作用。传统采用人工阅片的方式需要有经验的医生,这就增加了判断的主观性,容易产生误判的可能。为解决这一问题,提出了一种融合注意力机制模块的U-Net改进网络,使模型学习到更多的内容,可以更好地把训练过程集中在对应的病灶上。在BraTS公共数据集划分出的独立测试集上测试该模型,相关指标高于其他对比分割网络。
As a non-interventional diagnostic tool,MRI images of brain tumors contain extremely rich pathological,physio-logical and anatomical information,provide tremendous support for clinical diagnosis,and play an important role in medical re-search and disease treatment.The traditional method of manual reading requires experienced doctors,which increases the subjectivi-ty of judgment and is prone to misjudgment.In order to solve this problem,this paper proposes a U-Net improved network that inte-grates attention mechanism modules,so that the model can learn more content and can better focus the training process on the corre-sponding lesions.The model is tested on an independent test set divided by the BraTS public data set,and the relevant indicators are higher than other comparative segmentation networks.
作者
王皓
WANG Hao(School of Computer Science,Beijing University of Posts and Telecommunications,Beijing 100876)
出处
《计算机与数字工程》
2023年第12期2954-2958,共5页
Computer & Digital Engineering
