Abstract
Interactive medical image segmentation based on human-in-the-loop machine learning is a novel paradigm that draws on human expert knowledge to assist medical image segmentation. However, existing methods often fall into what we call interactive misunderstanding, the essence of which is the dilemma in trading off short- and long-term interaction information. To better use the interaction information at various timescales, we propose an interactive segmentation framework, called interactive MEdical image segmentation with self-adaptive Confidence CAlibration (MECCA), which combines action-based confidence learning and multi-agent reinforcement learning. A novel confidence network is learned by predicting the alignment level of the action with short-term interaction information. A confidence-based reward-shaping mechanism is then proposed to explicitly incorporate confidence in the policy gradient calculation, thus directly correcting the model’s interactive misunderstanding. MECCA also enables user-friendly interactions by reducing the interaction intensity and difficulty via label generation and interaction guidance, respectively. Numerical experiments on different segmentation tasks show that MECCA can significantly improve short- and long-term interaction information utilization efficiency with remarkably fewer labeled samples. The demo video is available at https://bit.ly/mecca-demo-video.
摘要
基于人机交互的医疗图像分割方法是一种新的范式, 其通过引入专家交互信息来指导算法完成图像分割任务。然而, 现有医疗图像分割模型往往容易产生“交互误解”, 即无法合理权衡短期和长期交互信息的重要性。为更好地利用不同时间尺度上的交互信息, 本文提出一种基于自适应置信度校准的交互式医疗图像分割框架MECCA, 其结合了基于分割决策的置信度学习技术和多智能体强化学习技术, 并通过预测分割决策与短期交互信息的对齐水平来学习一个新颖的置信度网络。随后, 提出一种基于置信度的奖励塑造机制, 在策略梯度计算中引入置信度, 从而直接纠正模型产生的交互误解。MECCA还通过标签生成和交互指导来降低交互强度和难度, 从而实现用户友好交互。实验结果表明, MECCA在不同分割任务中可以显著提高短期和长期交互信息的利用效率, 且仅需较少的标注样本。演示视频可通过https://bit.ly/mecca-demo-video访问。
This is a preview of subscription content, log in via an institution to check access.
Similar content being viewed by others
Data availability
The demo video is available at https://bit.ly/meccademo-video. The other data that support the findings of this study are available from the corresponding author upon reasonable request.
References
Abel D, Jinnai Y, Guo SY, et al., 2018. Policy and value transfer in lifelong reinforcement learning. Proc 35th Int Conf on Machine Learning, p.20–29.
Achanta R, Shaji A, Smith K, et al., 2012. SLIC superpixels compared to state-of-the-art superpixel methods. IEEE Trans Patt Anal Mach Intell, 34(11):2274–2282. https://doi.org/10.1109/TPAMI.2012.120
Acuna D, Ling H, Kar A, et al., 2018. Efficient interactive annotation of segmentation datasets with Polygon-RNN++. Proc IEEE/CVF Conf on Computer Vision and Pattern Recognition, p.859–868. https://doi.org/10.1109/CVPR.2018.00096
Aljabri M, AlAmir M, AlGhamdi M, et al., 2022. Towards a better understanding of annotation tools for medical imaging: a survey. Multim Tools Appl, 81(18):25877–25911. https://doi.org/10.1007/s11042-022-12100-1
Bredell G, Tanner C, Konukoglu E, 2018. Iterative interaction training for segmentation editing networks. Proc 9th Int Workshop on Machine Learning in Medical Imaging, p.363–370. https://doi.org/10.1007/978-3-030-00919-9_42
Castrejón L, Kundu K, Urtasun R, et al., 2017. Annotating object instances with a polygon-RNN. IEEE Conf on Computer Vision and Pattern Recognition, p.4485–4493. https://doi.org/10.1109/CVPR.2017.477
DeVries T, Taylor GW, 2018a. Learning confidence for out-of-distribution detection in neural networks. https://doi.org/10.48550/arXiv.1802.04865
DeVries T, Taylor GW, 2018b. Leveraging uncertainty estimates for predicting segmentation quality. https://doi.org/10.48550/arXiv.1807.00502
Feng RW, Zheng XS, Gao TX, et al., 2021. Interactive few-shot learning: limited supervision, better medical image segmentation. IEEE Trans Med Imag, 40(10):2575–2588. https://doi.org/10.1109/TMI.2021.3060551
Furuta R, Inoue N, Yamasaki T, 2020. PixelRL: fully convolutional network with reinforcement learning for image processing. IEEE Trans Multim, 22(7):1704–1719. https://doi.org/10.1109/TMM.2019.2960636
Glorot X, Bengio Y, 2010. Understanding the difficulty of training deep feedforward neural networks. Proc 13th Int Conf on Artificial Intelligence and Statistics, p.249–256.
Hung W, Tsai Y, Liou Y, et al., 2018. Adversarial learning for semi-supervised semantic segmentation. Proc British Machine Vision Conf, p.65.
Jungo A, Reyes M, 2019. Assessing reliability and challenges of uncertainty estimations for medical image segmentation. Proc 22nd Int Conf on Medical Image Computing and Computer-Assisted Intervention, p.48–56. https://doi.org/10.1007/978-3-030-32245-8_6
Kendall A, Gal Y, 2017. What uncertainties do we need in bayesian deep learning for computer vision? Proc 3rd Int Conf on Neural Information Processing System, p.5580–5590.
Kingma DP, Ba J, 2015. Adam: a method for stochastic optimization. Proc 3rd Int Conf on Learning Representations. https://doi.org/10.48550/arXiv.1412.6980
Lee KM, Song G, 2018. SeedNet: automatic seed generation with deep reinforcement learning for robust interactive segmentation. IEEE/CVF Conf on Computer Vision and Pattern Recognition, p.1760–1768. https://doi.org/10.1109/CVPR.2018.00189
Li L, Zimmer VA, Schnabel JA, et al., 2021. AtrialGeneral: domain generalization for left atrial segmentation of multi-center LGE MRIs. Proc 24th Int Conf on Medical Image Computing and Computer-Assisted Intervention, p.557–566. https://doi.org/10.1007/978-3-030-87231-1_54
Liao X, Li WH, Xu QS, et al., 2020. Iteratively-refined interactive 3D medical image segmentation with multiagent reinforcement learning. Proc IEEE/CVF Conf on Computer Vision and Pattern Recognition, p.9394–9402. https://doi.org/10.1109/CVPR42600.2020.00941
Lin D, Dai JF, Jia JY, et al., 2016. ScribbleSup: scribble-supervised convolutional networks for semantic segmentation. IEEE Conf on Computer Vision and Pattern Recognition, p.3159–3167. https://doi.org/10.1109/CVPR.2016.344
Lin TY, Goyal P, Girshick R, et al., 2017. Focal loss for dense object detection. Proc IEEE Int Conf on Computer Vision, p.2999–3007. https://doi.org/10.1109/ICCV.2017.324
Ma CF, Xu QS, Wang XF, et al., 2021. Boundary-aware supervoxel-level iteratively refined interactive 3D image segmentation with multi-agent reinforcement learning. IEEE Trans Med Imag, 40(10):2563–2574. https://doi.org/10.1109/TMI.2020.3048477
Menze BH, Jakab A, Bauer S, et al., 2015. The multimodal brain tumor image segmentation benchmark (BRATS). IEEE Trans Med Imag, 34(10):1993–2024. https://doi.org/10.1109/TMI.2014.2377694
Mnih V, Badia AP, Mirza M, et al., 2016. Asynchronous methods for deep reinforcement learning. Proc 33rd Int Conf on Machine Learning, p.1928–1937.
Moeskops P, Veta M, Lafarge MW, et al., 2017. Adversarial training and dilated convolutions for brain MRI segmentation. Proc 3rd Int Workshop on Deep Learning in Medical Image Analysis and 7th Int Workshop on Multimodal Learning for Clinical Decision Support, p.56–64. https://doi.org/10.1007/978-3-319-67558-9_7
Nie D, Wang L, Xiang L, et al., 2019. Difficulty-aware attention network with confidence learning for medical image segmentation. Proc 33rd AAAI Conf on Artificial Intelligence, 31st Innovative Applications of Artificial Intelligence Conf, and 9th AAAI Symp on Educational Advances in Artificial Intelligence, p.1085–1092. https://doi.org/10.1609/aaai.v33i01.33011085
OpenAI, 2022. ChatGPT: Optimizing Language Models for Dialogue. https://openai.casa/blog/chatgpt/ [Accessed on July 10, 2022].
Paszke A, Gross S, Massa F, et al., 2019. PyTorch: an imperative style, high-performance deep learning library. Proc 33rd Int Conf on Neural Information Processing Systems, p.8026–8037.
Prabhu A, Torr PHS, Dokania PK, 2020. GDumb: a simple approach that questions our progress in continual learning. Proc 16th European Conf on Computer Vision, p.524–540. https://doi.org/10.1007/978-3-030-58536-5_31
Rajchl M, Lee MCH, Oktay O, et al., 2017. DeepCut: object segmentation from bounding box annotations using convolutional neural networks. IEEE Trans Med Imag, 36(2):674–683. https://doi.org/10.1109/TMI.2016.2621185
Rebuffi SA, Kolesnikov A, Sperl G, 2017. iCaRL: incremental classifier and representation learning. IEEE Conf on Computer Vision and Pattern Recognition, p.5533–5542. https://doi.org/10.1109/CVPR.2017.587
Robinson R, Oktay O, Bai WJ, et al., 2018. Real-time prediction of segmentation quality. Proc 21st Int Conf on Medical Image Computing and Computer-Assisted Intervention, p.578–585. https://doi.org/10.1007/978-3-030-00937-3_66
Ronneberger O, Fischer P, Brox T, 2015. U-Net: convolutional networks for biomedical image segmentation. Proc 18th Int Conf on Medical Image Computing and Computer-Assisted Intervention, p.234–241. https://doi.org/10.1007/978-3-319-24574-4_28
Shrivastava A, Gupta A, Girshick R, 2016. Training region-based object detectors with online hard example mining. IEEE Conf on Computer Vision and Pattern Recognition, p.761–769. https://doi.org/10.1109/CVPR.2016.89
Simpson AL, Antonelli M, Bakas S, et al., 2019. A large annotated medical image dataset for the development and evaluation of segmentation algorithms. https://doi.org/10.48550/arXiv.1902.09063
Wang GT, Li WQ, Zuluaga MA, et al., 2018. Interactive medical image segmentation using deep learning with image-specific fine tuning. IEEE Trans Med Imag, 37(7):1562–1573. https://doi.org/10.1109/TMI.2018.2791721
Wang GT, Zuluaga MA, Li WQ, et al., 2019. DeepIGeoS: a deep interactive geodesic framework for medical image segmentation. IEEE Trans Patt Anal Mach Intell, 41(7):1559–1572. https://doi.org/10.1109/TPAMI.2018.2840695
Xie AN, Harrison J, Finn C, 2020. Deep reinforcement learning amidst lifelong non-stationarity. https://doi.org/10.48550/arXiv.2006.10701
Xu N, Price B, Cohen S, et al., 2016. Deep interactive object selection. IEEE Conf on Computer Vision and Pattern Recognition, p.373–381. https://doi.org/10.1109/CVPR.2016.47
Ye QH, Gao Y, Ding WP, et al., 2022. Robust weakly supervised learning for COVID-19 recognition using multicenter CT images. Appl Soft Comput, 116:108291. https://doi.org/10.1016/j.asoc.2021.108291
Yu LQ, Wang SJ, Li XM, et al., 2019. Uncertainty-aware self-ensembling model for semi-supervised 3D left atrium segmentation. Proc 22nd Int Conf on Medical Image Computing and Computer-Assisted Intervention, p.605–613. https://doi.org/10.1007/978-3-030-32245-8_67
Zhang KQ, Yang ZR, Basar T, 2021. Decentralized multiagent reinforcement learning with networked agents: recent advances. Front Inform Technol Electron Eng, 22(6):802–814. https://doi.org/10.1631/FITEE.1900661
Zhang SY, Liew JH, Wei YC, et al., 2020. Interactive object segmentation with inside-outside guidance. Proc IEEE/CVF Conf on Computer Vision and Pattern Recognition, p.12231–12241. https://doi.org/10.1109/CVPR42600.2020.01225
Zhuang XH, Shen J, 2016. Multi-scale patch and multi-modality atlases for whole heart segmentation of MRI. Med Image Anal, 31:77–87. https://doi.org/10.1016/j.media.2016.02.006
Author information
Authors and Affiliations
Contributions
Chuyun SHEN, Wenhao LI, and Qishen XU designed the research and conducted the experiments. Bin HU, Fengping ZHU, and Yuxin LI ensured the validity of the experiments. Bo JIN, Haibin CAI, and Xiangfeng WANG offered support across various experimental aspects. Chuyun SHEN drafted the paper. All the authors revised and finalized the paper.
Corresponding author
Ethics declarations
Chuyun SHEN, Wenhao LI, Qisen XU, Bin HU, Bo JIN, Haibin CAI, Fengping ZHU, Yuxin LI, and Xiangfeng WANG declare that they have no conflict of interest.
Additional information
Project supported by the Science and Technology Commission of Shanghai Municipality, China (No. 22511106004), the Postdoctoral Science Foundation of China (No. 2022M723039), the National Natural Science Foundation of China (No. 12071145), and the Shanghai Trusted Industry Internet Software Collaborative Innovation Center, China
List of supplementary materials
1 More related works
2 More visualizations
3 Robustness of MECCA
4 Comparison of baseline responses to the same user interaction
Fig. S1 MECCA segmentation process
Fig. S2 Qualitative segmentation results of MECCA for the BraTS2015 validation set
Figs. S3–S5 Results of different methods’ responses to the same user interactions according to the same initial segmentation on different testing instances and different channels for the Liver dataset in Medical Segmentation Decathlon
Table S1 Dice of our method which varies with the number of interactions under different cases
Table S2 MECCA’s tolerance to inaccurate interaction points
Supplementary materials
Rights and permissions
About this article
Cite this article
Shen, C., Li, W., Xu, Q. et al. Interactive medical image segmentation with self-adaptive confidence calibration. Front Inform Technol Electron Eng 24, 1332–1348 (2023). https://doi.org/10.1631/FITEE.2200299
Received:
Accepted:
Published:
Issue Date:
DOI: https://doi.org/10.1631/FITEE.2200299
Key words
- Medical image segmentation
- Interactive segmentation
- Multi-agent reinforcement learning
- Confidence learning
- Semi-supervised learning