文章摘要

文章信息/Info

Title:: Multi-scale edge optimization MRI segmentation algorithm for bladder cancer based on Swin Transformer

作者:: 刘力宾1; 2; 李翔2; 3; 魏本征2; 3; 张魁星1; 2△; （1.山东中医药大学智能与信息工程学院，济南 250355；2.山东中医药大学医学人工智能研究中心，青岛 266112；3.山东中医药大学青岛中医药科学院，青岛 266112）

Author(s):: LIU Libin1; 2; LI Xiang2; 3; WEI Benzheng2; 3; ZHANG Kuixing1; 2; （1.College of Intelligence and Information Engineering,Shandong University of Traditional Chinese Medicine, Jinan 250355, China；2.Center for Medical Artificial Intelligence,Shandong University of Traditional Chinese Medicine, Qingdao 266112, China;3. Qingdao Academy of Chinese Medical Sciences, Shandong University of Traditional Chinese Medicine, Qingdao 266112）

Keywords:: Bladder cancer; Magnetic resonance imaging; Swin Transformer; Multi-scale feature extraction; Segmented edge optimization

摘要:: 针对膀胱癌磁共振成像(magnetic resonance imaging, MRI)肿瘤区域小、癌变区域边缘模糊等问题，本研究设计了一种以Swin Transformer为骨干网络的多尺度边缘优化膀胱癌MRI分割算法。首先，设计特征提取模块学习细粒度语义特征信息；然后，采用一种多尺度特征金字塔模块解决肿瘤形状复杂多变的问题；最后，采用边缘细节解码模块对肿瘤分割边缘进行优化，提高分割性能。采用五折交叉实验验证算法性能，结果显示，该算法的交并比（IOU）、骰子系数（DICE）和准确率（ACC）分别达到89.11%、93.73%和93.46%。结果表明，本算法分割性能优良，可实现膀胱癌MRI影像精准分割，为临床医生提供辅助诊断工具。

Abstract:: Due to the small tumor area in magnetic resonance imaging（MRI）of bladder cancer and the blurred edge of the cancerous area, we designed a multi-scale feature extraction fusion algorithm for bladder cancer MRI image segmentation based on Swin transformer as the backbone network. Firstly, the feature extraction module was designed to learn fine-grained semantic feature information. A multi-scale feature pyramid module was used to solve the problem of complex and changeable tumor shapes. Finally, an edge detail decoding module was used to optimize the tumor segmentation edge and improve segmentation performance. The 5-fold cross was used to verify the algorithm performance. The results showed that the intersection over union (IOU), dice coefficient (DICE) and accuracy rate (ACC) of this algorithm were 89.11%, 93.73% and 93.46%, respectively.The experimental results indicate that this algorithm has excellent segmentation performance, which can realize accurate MRI image segmentation of bladder cancer and provide auxiliary diagnostic tools for clinicians.