基于 Mask RCNN 的桥小脑角区脑膜瘤与听神经瘤分类定位研究

doi:10.11938/cjmr20202825

波谱学杂志 ›› 2021, Vol. 38 ›› Issue (1): 58-68.doi: 10.11938/cjmr20202825

基于 Mask RCNN 的桥小脑角区脑膜瘤与听神经瘤分类定位研究

刘颖, 陈静聪, 胡小洋, 章浩伟

上海理工大学医疗器械与食品学院, 医学影像工程研究所, 上海 200093

收稿日期:2020-04-04 出版日期:2021-03-05 发布日期:2020-05-22
通讯作者: 刘颖 Tel: 18602168660, E-mail: ling2431@163.com. E-mail:ling2431@163.com
基金资助:
微创励志创新基金资助项目（182702156）.

Classification and Localization of Meningioma and Acoustic Neuroma in Cerebellopontine Angle Based on Mask RCNN

LIU Ying, CHEN Jing-cong, HU Xiao-yang, ZHANG Hao-wei

Institute of Medical Imaging Engineering, School of Medical Instrument and Food Engineering, University of Shanghai for Science and Technology, Shanghai 200093, China

Received:2020-04-04 Online:2021-03-05 Published:2020-05-22

摘要/Abstract

摘要： 由于人体桥小脑角区的脑膜瘤与听神经瘤在影像学的表现以及发病位置极其相似，所以临床诊断极易发生误诊．针对此问题，本文应用掩膜区域卷积神经网络（Mask RCNN）对两类肿瘤进行分类定位研究．首先采集89名脑膜瘤与218名听神经瘤患者的T₁WI-SE序列的磁共振图像，对其进行预处理，再结合改进的特征金字塔网络（FPN）算法进行网络训练．本文对比了三种不同的Mask RCNN主干网络对两者分类定位的效果．结果表明，结合改进的FPN算法和ResNet101作为主干网络的Mask RCNN分类定位模型能够有效实现对两类肿瘤的分类定位，精确率为0.918 2、召回率为0.856 9、特异性为0.876 2、均值平均精度（mAP）为0.90．

关键词: 掩膜区域卷积神经网络(Mask RCNN), 特征金字塔网络(FPN)算法, 分类定位, 脑膜瘤, 听神经瘤

Abstract: Differential diagnosis of meningioma and acoustic neuroma can be difficult because these two tumors have similar locations and appearances on medical images. To address this problem, mask region convolutional neural network (Mask RCNN) was used to classify and diagnose those two types of tumors. First, magnetic resonance images acquired with T₁-weighted spin-echo (T₁WI-SE) sequence of 89 meningioma and 218 acoustic neuroma patients were collected and preprocessed. Then the improved feature pyramid networks (FPN) algorithm was used for model network training. The effects of three different backbone feature extraction layers on classification and location were compared. It was demonstrated that Mask RCNN model with improved FPN and ResNet101 as backbone network is able to effectively classify and locate meningioma and acoustic neuroma, the values of precision, recall, specificity, and mean average precision (mAP) are 0.918 2, 0.856 9, 0.876 2, and 0.90, respectively.

Key words: Mask RCNN, FPN algorithm, classification and localization, meningioma, acoustic neuroma

中图分类号:

刘颖, 陈静聪, 胡小洋, 章浩伟. 基于 Mask RCNN 的桥小脑角区脑膜瘤与听神经瘤分类定位研究[J]. 波谱学杂志, 2021, 38(1): 58-68.

LIU Ying, CHEN Jing-cong, HU Xiao-yang, ZHANG Hao-wei. Classification and Localization of Meningioma and Acoustic Neuroma in Cerebellopontine Angle Based on Mask RCNN[J]. Chinese Journal of Magnetic Resonance, 2021, 38(1): 58-68.

参考文献

[1] NIU L, ZHOU X M, DUAN C F, et al. Differentiation researches on the meningioma subtypes by radiomics from contrast enhanced MRI, a preliminary study[J]. World Neurosurg, 2019, 126:e646-e652.
[2] YE D S, XU S H, HUANG Y C, et al. MRI diagnostic value of cerebellopontine angle tumors[J]. Chinese and Foreign Medical Research, 2017, 15(1):50-51.叶德湫,许淑惠,黄永础,等.桥小脑角区肿瘤的MRI诊断价值[J].中外医学研究, 2017, 15(1):50-51.
[3] MA Y, GUO H, WANG Q S, et al. Correlations between morphological characteristics and expression levels of specific molecular biomarkers in glioblastoma[J]. Chinese J Magn Reson, 2018, 35(1):22-30.马芸,郭虹,王秋实,等.基于影像的形态学特征与胶质母细胞瘤特征分子表达的相关性研究[J].波谱学杂志, 2018, 35(1):22-30.
[4] ZHANG Q. CT and MRI imaging analysis of occupying lesions in the cerebellar horn region[J]. Journal of Medical Imaging, 2019, 29(6):913-916.张芹.桥小脑角区占位性病变CT和MRI影像学分析[J].医学影像学杂志, 2019, 29(6):913-916.
[5] MA Y T, CHEN X Y, TIAN G S, et al. MRI characteristics of space-occupying lesions in the cerebellopontine angle region[J]. Chinese Clinical Research, 2017, 30(6):838-840.马云涛,陈秀艳,田贵森,等.桥小脑角区占位性病变核磁共振影像学特征[J].中国临床研究, 2017, 30(6):838-840.
[6] ZHANG M, ZHANG H Q. The application of MRI in the diagnosis of cerebellopontine angle tumor patients and its clinical value analysis[J]. China Medical Innovation, 2018, 15(34):25-29.张敏,张海青. MRI在桥小脑角区肿瘤患者病情诊断中的应用及其临床价值分析[J].中国医学创新, 2018, 15(34):25-29.
[7] XU N N, GE Y R, WANG J Y. Medical image recognition of brain MRI based on NSCT and SVM[J]. Modern Electronic Technology, 2014, 37(12):63-66+69.徐楠楠,葛玉荣,王佳奕.基于NSCT和SVM的脑MRI医学图像识别[J].现代电子技术, 2014, 37(12):63-66+69.
[8] LAVANYADEVI R, MACHAKOWSALYA M, NIVETHITHA J, et al. Brain tumor classification and segmentation in MRI images using PNN[C]. IEEE International Conference on Electrical. IEEE, 2017.
[9] LI Z N, DONG M H, WEN S P, et al. CLU-CNNs:Object detection for medical images[J]. Neurocomputing, 2019, 350:53-59.
[10] YANG X, LIU C Y, WANG Z W, et al. Co-trained convolutional neural networks for automated detection of prostate cancer in multi-parametric MRI[J]. Med Image Anal, 2017, 42:212-227.
[11] FELDMAN A M, DAI Z, CARVER E, et al. Utilizing a deep learning-based object detection and instance segmentation algorithm for the delineation of prostate and prostate cancer segmentation[J]. Int J Radiat Oncol, 2019, 105(1):S197-S198.
[12] OU P, LU K, ZHANG Z, et al. Object recognition and spatial localization based on Mask RCNN[J]. Computer Measurement & Control, 2019, 27(6):172-176.欧攀,路奎,张正,等.基于Mask RCNN的目标识别与空间定位[J].计算机测量与控制, 2019, 27(6):172-176.
[13] ZHANG B, ZHOU J, WANG F, et al. Detection and recognition of touch screen glass defects based on Mask R-CNN[J]. Software Guide, 2019, 18(2):64-67, 71.张博,周军,王芳,等.基于Mask R-CNN的触摸屏玻璃疵病检测与识别[J].软件导刊, 2019, 18(2):64-67, 71.
[14] WANG F, LU J D, SHEN G R, et al. Image segmentation of green apple based on CLAHE and open-close operation[J]. Computer Measurement and Control, 2017, 25(2):141-145.王帆,吕继东,申根荣,等.基于CLAHE和开闭运算的绿色苹果图像分割[J].计算机测量与控制, 2017, 25(2):141-145.
[15] LI L, LI W J, WU Y Z. Underwater image enhancement algorithm based on red dark channel prior theory and CLAHE algorithm[J]. China Ship Research, 2019, 14(S1):175-182.李炼,李维嘉,吴耀中.基于红色暗通道先验理论与CLAHE算法的水下图像增强算法[J].中国舰船研究, 2019, 14(S1):175-182.
[16] BAI T, PANG Y, WANG J C, et al. An optimized faster R-CNN method based on DRNet and ROI align for building detection in remote sensing images[J]. Remote Sens, 2020, 12(5):762.
[17] HE K M, GKIOXARI G, DOLLÁR P, et al. Mask R-CNN[C]. Proceedings of the IEEE international conference on computer vision. 2017:2961-2969.
[18] MALHOTR K R, DAVOUDI A, SIEGEL S, et al. Autonomous detection of disruptions in the intensive care unit using deep mask R-CNN[C]. IEEE/CVF Conference on Computer Vision & Pattern Recognition Workshops. IEEE Computer Society, 2018.
[19] REN Z J, YUN S Z, LI D W, et al. Mask R-CNN target detection method based on improved feature pyramid[J]. Laser & Optoelectronics Progress, 2019, 56(4):174-179.任之俊,蔺素珍,李大威,等.基于改进特征金字塔的Mask R-CNN目标检测方法[J].激光与光电子学进展, 2019, 56(4):174-179.
[20] CHANG S, ZHANG F, HUANG S, et al. Siamese feature pyramid network for visual tracking[C]. 2019 IEEE/CIC International Conference on Communications Workshops in China (ICCC Workshops). IEEE, 2019.
[21] 黄泽桑.基于深度学习的目标检测技术研究[D].北京:北京邮电大学, 2019.
[22] WANG H Z, ZHAO D, YANG L Q, et al. An approach for training data enrichment and batch labeling in AI+MRI aided diagnosis[J]. Chinese J Magn Reson, 2018, 35(4):50-59.汪红志,赵地,杨丽琴,等.基于AI+MRI的影像诊断的样本增广与批量标注方法[J].波谱学杂志, 2018, 35(4):50-59.

基于 Mask RCNN 的桥小脑角区脑膜瘤与听神经瘤分类定位研究

Classification and Localization of Meningioma and Acoustic Neuroma in Cerebellopontine Angle Based on Mask RCNN

PDF

可视化

摘要/Abstract

引用本文

使用本文

参考文献

相关文章 1

编辑推荐

Metrics

本文评价