Current Issue Cover

张蝶1, 黄慧1, 马燕1, 黄丙仓2, 陆炜平2(1.上海师范大学信息与机电工程学院;2.上海市浦东新区公利医院影像科)

摘 要
目的 前列腺图像的精确分割对患者健康状况的评估以及后续治疗方案的设计有重大意义。目前MRI图像普遍存在组织与器官之间对比度低、边界模糊和噪声大等特点,给MRI前列腺图像分割带来极大的挑战。为此,本文提出一种基于边缘信息和通道注意力机制的2D前列腺分割模型,增强组织之间的边缘信息、降低图像噪声影响,从而达到提高前列腺分割效果的目的。方法 该模型通过对标准U-Net架构进行修改,将普通卷积替换为深度可分离卷积,达到缓解模型过拟合的目的,通过ECA注意力机制对U-Net解码器特征进行优化,以放大并保存小尺度目标的信息,又提出边缘信息模块和边缘信息金字塔模块,以缓解频繁下采样带来的边缘信息衰退以及编码器和解码器特征之间的语义差距的问题。结果 最后,我们利用PROMISE12数据集来验证模型的有效性,并与6种基于U-Net的图像分割方法进行对比,实验证明其分割结果在Dice系数(Dice coefficient,DC)、召回率(Recall)、Jaccard系数(Jac)和准确度(Accuracy)等指标上均有提高,DC较U-Net提高了8.81%,Jac较U-Net++和Attention U-Net分别提高了3.42%和6.86%。结论 本文提出的基于边缘信息和通道注意力机制的U-Net模型(AMF-U-Net)生成的分割图像具有丰富的边缘信息和空间信息,其主观效果和客观评价指标均优于其他同类方法,为提高临床诊断的准确度提供帮助。
Prostate segmentation model based on channel attention mechanism and marginal information fusion

Zhang Die, Huang Hui1, Ma Yan2, Huang Bingcang2, Lu Weiping2(1.The College of Information,Mechanical and Electrical Engineering,Shanghai Normal University;2.Department of Radiology,Gongli Hospital of Shanghai Pudong New Area)

Objective Prostate cancer, which is an epithelial malignancy that occurs in the prostate, is one of the most common malignant diseases. Early detection of potentially cancerous prostate is important to reduce prostate cancer mortality. Magnetic Resonance Imaging (MRI) is one of the most commonly used imaging methods for the detection of prostate in clinical practice, commonly used for the detection, localization and segmentation of prostate cancer. It is significant for doctors to formulate suitable medical plan for patients and postoperative record. In computer-aided diagnosis, it is often necessary to extract the prostate region from the image, and further calculate the corresponding characteristics for physiological analysis and pathological research, in order to assist clinicians to make accurate judgments. Current methods of MRI prostate segmentation can be divided into two categories: traditional methods and deep learning-based methods. The traditional segmentation method is based on the analysis of the features extracted from the image with knowledge to image processing. The effect of this kind of method depends on the performance of the extracted features, and sometimes requires more manual interaction. In recent years, with the continuous development of computer technology, deep learning technology has been widely applied to image segmentation. Different from visible light images, medical images have some special characteristics: