基于深度学习的高分辨率食管测压图谱中食管收缩活力分类

贺福利; 戴渝卓; 李钊颖; 粟日; 曹聪; 王姣菊; 戴燎元; 侯木舟; 汪政

doi:10.11999/JEIT210909

基于深度学习的高分辨率食管测压图谱中食管收缩活力分类

doi: 10.11999/JEIT210909

贺福利¹,
戴渝卓¹,
李钊颖¹,
粟日¹,
曹聪¹,
王姣菊¹,
戴燎元³,
侯木舟^1, ,,
汪政^{1, 2}

1.
中南大学数学与统计学院长沙 410083
2.
湖南第一师范大学科学与工程学院长沙 410205
3.
湖南省计算机用户协会长沙 410205

基金项目: 中南大学研究生创新研究基金(2020zzts362)，湖南省自然科学基金(2020JJ4105)

详细信息

作者简介:
贺福利：男，1981年生，副教授，硕士研究生导师，研究方向为复分析、Clifford分析、Riemann Hilbert问题及其相关、数学建模与应用

戴渝卓：女，1997年生，硕士生，研究方向为医疗大数据、医疗图像处理、推荐系统、信息探索、人工智能算法

曹聪：男，1993年生，博士生，研究方向为医疗大数据、人工智能算法、统计优化、机器学习

王姣菊：女，1995年生，博士生，研究方向为医疗图像处理、深度学习、生成式对抗网络

侯木舟：男，1965年生，教授，博士研究生导师，研究方向为区块链技术及应用、医学临床大数据分析，图像与异常信号的分析与检测

汪政：男，1975年生，博士生，研究方向为计算机视觉与图像处理

通讯作者:
侯木舟　houmuzhou@sina.com

中图分类号: TN911.73
计量
- 文章访问数: 668
- HTML全文浏览量: 374
- PDF下载量: 87
- 被引次数: 0
出版历程
- 收稿日期: 2021-08-31
- 修回日期: 2021-11-30
- 录用日期: 2021-12-20
- 网络出版日期: 2021-12-27
- 刊出日期: 2022-01-10

Deep Learned Esophageal Contraction Vigor Classification on High-resolution Manometry Images

1.
School of Mathematics and Statistics, Central South University, Changsha 410083, China
2.
Science and Engineering School, Hunan First Normal University, Changsha 410205, China
3.
Hunan Computer Users Association, Changsha 410205, China

Funds: Graduate Student Innovation Foundation of Central South University (2020zzts362), The Natural Science Foundation of Hunan Province (2020JJ4105)

摘要

摘要: 高分辨率食管测压技术(HRM)作为检测食管动力障碍性疾病(EMD)的金标准，已广泛应用于临床试验以辅助医生进行诊断治疗。随着患病率的上升，HRM图像的数据量爆炸式增长，加之EMD的诊断流程较为复杂，临床上EMD误诊事件时有发生。为了提高EMD诊断的准确性，希望搭建一个计算机辅助诊断(Computer Aided Diagnosis, CAD)系统帮助医生对HRM图像进行自动分析。由于食管收缩活力的异常是诊断EMD的重要依据，该文提出了一个深度学习模型(PoS-ClasNet)以完成对HRM图像的食管收缩活力分类任务，为今后机器代替人工诊断EMD奠定基础。PoS-ClasNet作为一个多任务卷积神经网络(CNN)由PoSNet和S-ClasNet构成。前者用于HRM图像中吞咽框的检测和提取任务，后者根据食管吞咽特征鉴别收缩活力类型。实验使用了4000幅专家标记的HRM图像，用于训练、验证和测试的图像分别占比为70%, 20%和10%。在测试集上，食管收缩活力分类器PoS-ClasNet的分类准确率高达93.25%，精度和召回率分别为93.39%和93.60%。结果表明PoS-ClasNet能较好地适应HRM图像数据的特性，在智能诊断食管收缩活力的任务中表现出了不俗的准确性和稳健性。将它应用在临床上辅助医生诊疗，会带来巨大的社会效益。
- 高分辨率食管测压 /
- 食管收缩活力 /
- 深度学习 /
- 卷积神经网络
Abstract: As the gold standard for the detection of Esophageal Motility Disorder(EMD), High-Resolution Manometry(HRM) is widely used in clinical tests to assist doctors in diagnosis. The amount of HRM images explodes with an increase in the prevalence rate, and the diagnostic process of EMD is complicated, both of which may lead to misdiagnosis of EMD in clinic. To improve the accuracy of the diagnosis of EMD, we hope to build a Computer Aided Diagnosis(CAD) system to assist doctors in analyzing HRM images automatically. Since the abnormality of esophageal contraction vigor is an important basis for diagnosis of EMD, in this paper, a Deep Learning(DL) model(PoS-ClasNet) is proposed to classify esophageal contraction vigor, which lays the foundation for machine to diagnose EMD instead of manual in the future. PoS-ClasNet, as a multi-task Convolutional Neural Network(CNN), is formed by PoSNet and S-ClassNet. The former is used to detect and extract swallowing frames in HRM images, while the latter identifies the type of contraction vigor based on esophageal swallowing characteristics. 4,000 expert-labeled HRM images are used for the experiment, among which the images of training set, verification set and test set accounted for 70%, 20% and 10%. On the test set, the classification accuracy of esophageal contraction vigor classifier PoS-ClasNet is as high as 93.25%, meanwhile the precision rate and the recall rate are 93.39% and 93.60% respectively. The experimental results show PoS-ClasNet can well adapt to the features of HRM image, with the outstanding accuracy and robustness in the task of intelligent diagnosis of esophageal contraction vigor. If the proposed model is used to assist doctors in clinical prevention, diagnosis and treatment, it will bring enormous social benefits.
- High-Resolution Manometry (HRM) /
- Esophageal contraction vigor /
- Deep Learning (DL) /
- Convolutional Neural Network (CNN)

HTML全文

图 1 高分辨率食管测压图谱数据集

下载: 全尺寸图片幻灯片

图 2 食管收缩活力类型

下载: 全尺寸图片幻灯片

图 3 受噪声污染的HRM图像

下载: 全尺寸图片幻灯片

图 4 预处理前后的HRM图像

下载: 全尺寸图片幻灯片

图 5 PoS-ClasNet的模型结构图

下载: 全尺寸图片幻灯片

图 6 所有分类模型的分类准确率柱状图

下载: 全尺寸图片幻灯片

图 7 各网络模型的ROC曲线

下载: 全尺寸图片幻灯片

图 8 S-ClasNet的学习曲线

下载: 全尺寸图片幻灯片

图 9 测试集上的3类收缩活力分类的ROC曲线

下载: 全尺寸图片幻灯片

图 10 测试集上的混淆矩阵

下载: 全尺寸图片幻灯片

表 1 不同分类模型对食管收缩活力分类的结果比较(均值±标准差)

		Inceptionresnetv2	Inceptionv3	Resnet50	Xception	S-ClasNet
损失	训练集	0.2284±0.0830	0.2784±0.0890	0.3317±0.0806	0.2514±0.0641	0.2574±0.0574
损失	验证集	0.2480±0.0535	0.2784±0.0940	0.3203±0.0825	0.2524±0.0421	0.2503±0.0429
准确率	训练集	0.9148±0.0339	0.8965±0.0376	0.8733±0.0318	0.9056±0.0278	0.9015±0.0249
准确率	验证集	0.9138±0.0247	0.8988±0.0524	0.8945±0.0248	0.9157±0.0169	0.9122±0.0162
精度	训练集	0.9185±0.0311	0.9009±0.0338	0.8776±0.0299	0.9093±0.0253	0.9040±0.0250
精度	验证集	0.9162±0.0247	0.9023±0.0520	0.8990±0.0219	0.9191±0.0159	0.9143±0.0164
召回率/敏感性	训练集	0.9102±0.0391	0.8906±0.0440	0.8673±0.0346	0.9004±0.0332	0.8980±0.0294
召回率/敏感性	验证集	0.9112±0.0246	0.8948±0.0542	0.8911±0.0278	0.9121±0.0175	0.9104±0.0167
F1得分	训练集	0.9137±0.0358	0.8949±0.0400	0.8717±0.0326	0.9041±0.0300	0.9005±0.0278
F1得分	验证集	0.9133±0.0246	0.8980±0.0532	0.8944±0.0248	0.9151±0.0167	0.9121±0.0165
特异性	训练集	0.9599±0.0143	0.9515±0.0149	0.9397±0.0140	0.9554±0.0112	0.9527±0.0106
特异性	验证集	0.9584±0.0124	0.9516±0.0254	0.9500±0.0107	0.9599±0.0079	0.9573±0.0082

下载: 导出CSV

表 2 S-ClasNet在测试集上对食管收缩活力分类的结果比较

收缩活力分类	评判指标
收缩活力分类	准确率(%)	精度(%)	召回率(%)	F1得分
正常收缩	97.25	96.48	95.80	0.9614
弱收缩	93.25	90.37	89.71	0.9004
失收缩	96.00	92.68	94.21	0.9344
S-ClasNet	93.25	93.18	93.24	0.9321

下载: 导出CSV

参考文献(32)

[1]	BOWERS S P. Esophageal motility disorders[J]. Surgical Clinics of North America, 2015, 95(3): 467–482. doi: 10.1016/j.suc.2015.02.003
[2]	ZAMBITO G, ROETHER R, KERN B, et al. Is barium esophagram enough? Comparison of esophageal motility found on barium esophagram to high resolution manometry[J]. The American Journal of Surgery, 2021, 221(3): 575–577. doi: 10.1016/j.amjsurg.2020.11.028
[3]	KUNIEDA K, FUJISHIMA I, WAKABAYASHI H, et al. Relationship between tongue pressure and pharyngeal function assessed using high-resolution manometry in older dysphagia patients with sarcopenia: A pilot study[J]. Dysphagia, 2021, 36(1): 33–40. doi: 10.1007/s00455-020-10095-1
[4]	李飞, 王美峰, 汤玉蓉, 等. 《食管动力障碍的测压(第4版芝加哥分类)》更新点解读[J]. 中华消化杂志, 2021, 41(7): 492–497. doi: 10.3760/cma.j.cn311367-20210323-00173
[5]	PANDOLFINO J E, FOX M R, BREDENOORD A J, et al. High-resolution manometry in clinical practice: Utilizing pressure topography to classify oesophageal motility abnormalities[J]. Neurogastroenterology & Motility, 2009, 21(8): 796–806. doi: 10.1111/j.1365-2982.2009.01311.x
[6]	SAWADA A, GUZMAN M, NIKAKI K, et al. Identification of different phenotypes of esophageal reflux hypersensitivity and implications for treatment[J]. Clinical Gastroenterology and Hepatology, 2021, 19(4): 690–698.E2. doi: 10.1016/j.cgh.2020.03.063
[7]	KAHRILAS P J, BREDENOORD A J, FOX M, et al. The Chicago Classification of esophageal motility disorders, v3.0[J]. Neurogastroenterology & Motility, 2015, 27(2): 160–174. doi: 10.1111/nmo.12477
[8]	GYAWALI C P. High resolution manometry: The Ray Clouse legacy[J]. Neurogastroenterology & Motility, 2012, 24(S1): 2–4. doi: 10.1111/j.1365-2982.2011.01836.x
[9]	SCHMIDHUBER J. Deep learning in neural networks[J]. Neural Networks, 2015, 61: 85–117. doi: 10.1016/j.neunet.2014.09.003
[10]	LECUN Y, BENGIO Y, and HINTON G. Deep learning[J]. Nature, 2015, 521(7553): 436–444. doi: 10.1038/nature14539
[11]	KOU Wenjun, CARLSON D A, BAUMANN A J, et al. A deep-learning-based unsupervised model on esophageal manometry using variational autoencoder[J]. Artificial Intelligence in Medicine, 2021, 112: 102006. doi: 10.1016/j.artmed.2020.102006
[12]	CARLSON D A, KOU Wenjun, ROONEY K P, et al. Achalasia subtypes can be identified with functional luminal imaging probe (FLIP) panometry using a supervised machine learning process[J]. Neurogastroenterology & Motility, 2021, 33(3): e13932. doi: 10.1111/nmo.13932
[13]	侯晓华. 消化道高分辨率测压图谱[M]. 北京: 科学出版社, 2014. HOU Xiaohua. High Resolution Manometry in Digestive Tract[M]. Beijing: Science Press, 2014.
[14]	RENA Y. High-resolution esophageal manometry: Interpretation in clinical practice[J]. Current Opinion in Gastroenterology, 2017, 33(4): 301–309. doi: 10.1097/MOG.0000000000000369
[15]	BREDENOORD A J and SMOUT A J. High-resolution manometry of the esophagus: More than a colorful view on esophageal motility?[J]. Expert Review of Gastroenterology & Hepatology, 2007, 1(1): 61–69. doi: 10.1586/17474124.1.1.61
[16]	LI Zewen, LIU Fan, YANG Wenjie, et al. A survey of convolutional neural networks: Analysis, applications, and prospects[J]. IEEE Transactions on Neural Networks and Learning Systems, To be published.
[17]	GU Jiuxiang, WANG Zhenhua, KUEN J, et al. Recent advances in convolutional neural networks[J]. Pattern Recognition, 2018, 77: 534–377. doi: 10.1016/j.patcog.2017.10.013
[18]	SARVAMANGALA D R and KULKARNI R V. Convolutional neural networks in medical image understanding: A survey[J]. Evolutionary Intelligence, 2021(4): 1–22. doi: 10.1007/s12065-020-00540-3
[19]	MOHAPATRA S, SWARNKAR T, and DAS J. Deep Convolutional Neural Network in Medical Image Processing[M]. BALAS V E, MISHRA B K, and KUMAR R. Handbook of Deep Learning in Biomedical Engineering. Amsterdam: Academic Press, 2021: 25–60.
[20]	LITJENS G, KOOI T, BEJNORDI B E, et al. A survey on deep learning in medical image analysis[J]. Medical Image Analysis, 2017, 42: 60–88. doi: 10.1016/j.media.2017.07.005
[21]	GREENSPAN H, VAN GINNEKEN B, and SUMMERS R M. Guest editorial deep learning in medical imaging: Overview and future promise of an exciting new technique[J]. IEEE Transactions on Medical Imaging, 2016, 35(5): 1153–1159. doi: 10.1109/TMI.2016.2553401
[22]	JANOWCZYK A and MADABHUSHI A. Deep learning for digital pathology image analysis: A comprehensive tutorial with selected use cases[J]. Journal of Pathology Informatics, 2016, 7: 29. doi: 10.4103/2153-3539.186902
[23]	RONNEBERGER O, FISCHER P, and BROX T. U-Net: Convolutional networks for biomedical image segmentation[C]. Proceedings of the 18th International Conference on Medical Image Computing and Computer-Assisted Intervention, Munich, Germany, 2015: 234–241.
[24]	SIMONYAN K and ZISSERMAN A. Very deep convolutional networks for large-scale image recognition[J]. arXiv: 1409.1556, 2014.
[25]	GIRSHICK R. Fast R-CNN[C]. Proceedings of 2015 IEEE International Conference on Computer Vision, Santiago, Chile, 2015: 1440–1448.
[26]	REN Shaoqing, HE Kaiming, GIRSHICK R, et al. Faster R-CNN: Towards real-time object detection with region proposal networks[J]. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2017, 39(6): 1137–1149. doi: 10.1109/TPAMI.2016.2577031
[27]	ZHANG Ziang, WU Chengdong, COLEMAN S, et al. DENSE-INception U-net for medical image segmentation[J]. Computer Methods and Programs in Biomedicine, 2020, 192: 105395. doi: 10.1016/j.cmpb.2020.105395
[28]	WANG Zheng, MENG Yu, WENG Futian, et al. An effective CNN method for fully automated segmenting subcutaneous and visceral adipose tissue on CT scans[J]. Annals of Biomedical Engineering, 2020, 48(1): 312–328. doi: 10.1007/s10439-019-02349-3
[29]	DROZDZAL M, VORONTSOV E, CHARTRAND G, et al. The importance of skip connections in biomedical image segmentation[C]. Proceedings of the 1st International Workshop on Large-Scale Annotation of Biomedical Data and Expert Label Synthesis, Athens, Greece, 2016: 179–187.
[30]	IOFFE S and SZEGEDY C. Batch normalization: Accelerating deep network training by reducing internal covariate shift[C]. Proceedings of the 32nd International Conference on International Conference on Machine Learning, Lille, France, 2015: 448–456.
[31]	HE Kaiming, ZHANG Xiangyu, REN Shaoqing, et al. Delving deep into rectifiers: Surpassing human-level performance on ImageNet classification[C]. Proceedings of 2015 IEEE International Conference on Computer Vision, Santiago, Chile, 2015: 1026–1034.
[32]	KINGMA D P and BA J. Adam: A method for stochastic optimization[C]. Proceedings of the 3rd International Conference on Learning Representations, San Diego, USA, 2014.