改进的M2det内窥镜息肉检测方法

doi:10.3778/j.issn.1002-8331.2008-0084

摘要/Abstract

摘要： 结直肠癌是一种致命的疾病，作为息肉的肠腺瘤被认为是结直肠癌的主要病因，因此在临床诊断中发现肠息肉是一项至关重要的任务。息肉检测通常由医生操作内窥镜来实现，由于肠道环境复杂，息肉影像数据量大，小尺度息肉不易辨识，息肉检查过程除了极其依赖医生经验之外，工作压力和强度也给医生带来了极大的负担，因此需要借助计算机辅助诊断技术来检测息肉，该技术可以有效地处理大量的息肉影像数据、发现早期息肉、提高息肉检测的准确率。目前的一些方法对小型息肉存在漏检，因此提出了一种改进的M2det方法用于息肉检测，通过FFMs模块融合主干网络特征，使图像特征得到了充分利用，在SFAM模块中增加scSENet注意力机制，保留了有效特征，抑制无用特征，采用Focal loss计算分类损失，解决了正负样本不平衡问题。大量实验表明，该方法可以有效地检测出息肉且优于前沿的息肉检测方法，在CVC15数据集上mAP、F1-score、F2-score分别提升到了98.25%、97.30%、97.98%。

关键词: 息肉检测, 计算机辅助诊断, M2det, 特征融合模块（FFMs）, 空间和通道上的压缩激励网络（scSENet）, Focal loss

Abstract: Colorectal cancer is a fatal disease. Intestinal adenoma, which is a polyp, is considered to be the main cause of colorectal cancer. Therefore, finding intestinal polyps in clinical diagnosis is a vital task. Polyp detection is usually achieved by doctors operating an endoscope. Due to the complex intestinal environment, the large amount of polyp imaging data, and the difficulty of identifying small-scale polyps, the polyp examination process is extremely dependent on the doctor’s experience, and the work pressure and intensity also bring the doctors. Therefore, it is necessary to use CAD to detect polyps. This technology can effectively process a large amount of polyp imaging data, find early polyps, and improve the accuracy of polyp detection. Some current methods fail to detect small polyps. Therefore, it proposes an improved M2det method for polyp detection. First, the FFMs module is used to fuse the backbone network features to make full use of the image features. Secondly, scSENet is added to the SFAM module. The attention mechanism retains effective features and suppresses useless features. Finally, Focal loss is used to calculate the classification loss, which solves the problem of imbalance between positive and negative samples. A large number of experiments show that this method can effectively detect polyps and is better than cutting-edge polyp detection methods. On the CVC15 data set, mAP, F1-score, and F2-score have been increased to 98.25%, 97.30%, and 97.98%, respectively.

Key words: polyp detection, computer-aided diagnosis, M2det, feature fusion module（FFMs）, spatial and channel squeeze-and-excitation networks（scSENet）, Focal loss

王博, 张丽媛, 师为礼, 杨华民, 蒋振刚. 改进的M2det内窥镜息肉检测方法[J]. 计算机工程与应用, 2022, 58(2): 193-200.

WANG Bo, ZHANG Liyuan, SHI Weili, YANG Huamin, JIANG Zhengang. Improved M2det Endoscopic Polyp Detection Method[J]. Computer Engineering and Applications, 2022, 58(2): 193-200.

参考文献

[1] 龙思丹，孙希珍，赵栋燕，等.肠息肉病因学相关性研究进展[J].医学综述，2020（14）：2728-2732.
LONG S D，SUN X Z，ZHAO D Y，et al.Research progress in etiological correlation of intestinal polyps[J].Medical Recapitulate，2020（14）：2728-2732.
[2] 程时丹，陈丽娜，吴巍，等.形似良性病变的早期胃癌的临床研究[J].胃肠病学和肝病学杂志，2008，17（3）：197-200.
CHENG S D，CHEN L N，WU W，et al.The clinical research of early gastric cancer with benign appearance[J].Chinese Journal of Gastroenterology and Hepatology，2008，17（3）：197-200.
[3] KRIZHEVSKY A，SUTSKEVER I，HINTON G E.Imagenet classification with deep convolutional neural networks[C]//Advances in Neural Information Processing Systems，2012：1097-1105.
[4] SHEN X，SUN K，ZHANG S，et al.Lesion detection of electronic gastroscope images based on multiscale texture feature[C]//2012 IEEE International Conference on Signal Processing，Communication and Computing（ICSPCC），2012：756-759.
[5] TAJBAKHSH N，GURUDU S R，LIANG J.Automated polyp detection in colonoscopy videos using shape and context information[J].IEEE Transactions on Medical Imaging，2016，35（2）：630-644.
[6] YUAN Y，QIN W，IBRAGIMOV B，et al.RIIS-DenseNet：rotation-invariant and image similarity constrained densely connected convolutional network for polyp detection[C]//Medical Image Computing and Computer Assisted Intervention-MICCAI 2018.Cham：Springer，2018.
[7] MO X，TAO K，WANG Q，et al.An efficient approach for polyps detection in endoscopic videos based on faster R-CNN[C]//2018 24th International Conference on Pattern Recognition（ICPR），2018：3929-3934.
[8] GIRSHICK R.Fast R-CNN[C]//2015 IEEE International Conference on Computer Vision（ICCV），2016.
[9] REDMON J，DIVVALA S，GIRSHICK R，et al.You only look once：unified，real-time object detection[C]//Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition，2016：779-788.
[10] REN S，HE K，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.
[11] ZHAO Q，SHENG T，WANG Y，et al.M2det：a single-shot object detector based on multi-level feature pyramid network[C]//Proceedings of the AAAI Conference on Artificial Intelligence，2019：9259-9266.
[12] HU J，SHEN L，ALBANIE S，et al.Squeeze-and-excitation networks[J].IEEE Transactions on Pattern Analysis and Machine Intelligence，2017.
[13] ROY A G，NAVAB N，WACHINGER C.Concurrent spatial and channel squeeze & excitation in fully convolutional networks[C]//Proceedings of MICCAI，2018.
[14] MACLAURIN D，DUVENAUD D，ADAMS R P.Early stopping is nonparametric variational inference[J].arXiv：1504.01344，2015.
[15] LIN T Y，GOYAL P，GIRSHICK R，et al.Focal loss for dense object detection[J].IEEE Transactions on Pattern Analysis & Machine Intelligence，2020，42（2）：318-327.
[16] MISRA D.Mish：a self regularized non-monotonic neural activation function[J].arXiv：1908.08681，2019.
[17] SIMONYAN K，ZISSERMAN A.Very deep convolutional networks for large-scale image recognition[J].arXiv：1409.
1556，2014.
[18] VáZQUEZ D，BERNAL J，SáNCHEZ F J，et al.A benchmark for endoluminal scene segmentation of colonoscopy images[J].Journal of Healthcare Engineering，2017：4037190.
[19] KINGMA D P，BA J.Adam：a method for stochastic optimization[J].arXiv：1412.6980，2014.
[20] NEUBECK A，VAN GOOL L.Efficient non-maximum suppression[C]//18th International Conference on Pattern Recognition（ICPR’06），2006：850-855.
[21] BERNAL J，TAJKBAKSH N，SáNCHEZ F J，et al.Comparative validation of polyp detection methods in video colonoscopy：results from the MICCAI 2015 endoscopic vision challenge[J].IEEE Transactions on Medical Imaging，2017，36（6）：1231-1249.
[22] LI J，ZHANG J，CHANG D，et al.Computer-assisted detection of colonic polyps using improved faster R-CNN[J].Chinese Journal of Electronics，2019，28（4）：718-724.