This website requires JavaScript.

Parotid Gland MRI Segmentation Based on Swin-Unet and Multimodal Images

Yin DaiZi'an XuFayu Liu ...+3 Jun Fu
Jun 2022
摘要
Parotid gland tumors account for approximately 2% to 10% of head and necktumors. Preoperative tumor localization, differential diagnosis, and subsequentselection of appropriate treatment for parotid gland tumors is critical.However, the relative rarity of these tumors and the highly dispersed tissuetypes have left an unmet need for a subtle differential diagnosis of suchneoplastic lesions based on preoperative radiomics. Recently, deep learningmethods have developed rapidly, especially Transformer beats the traditionalconvolutional neural network in computer vision. Many new Transformer-basednetworks have been proposed for computer vision tasks. In this study,multicenter multimodal parotid gland MRI images were collected. The Swin-Unetwhich was based on Transformer was used. MRI images of STIR, T1 and T2modalities were combined into a three-channel data to train the network. Weachieved segmentation of the region of interest for parotid gland and tumor.The DSC of the model on the test set was 88.63%, MPA was 99.31%, MIoU was83.99%, and HD was 3.04. Then a series of comparison experiments were designedin this paper to further validate the segmentation performance of thealgorithm.
展开全部
图表提取

暂无人提供速读十问回答

论文十问由沈向洋博士提出,鼓励大家带着这十个问题去阅读论文,用有用的信息构建认知模型。写出自己的十问回答,还有机会在当前页面展示哦。

Q1论文试图解决什么问题?
Q2这是否是一个新的问题?
Q3这篇文章要验证一个什么科学假设?
0
被引用
笔记
问答