Pub Date : 2022-06-06DOI: 10.48550/arXiv.2206.02307
Chenyu You, Weichen Dai, L. Staib, J. Duncan
Contrastive learning has shown great promise over annotation scarcity problems in the context of medical image segmentation. Existing approaches typically assume a balanced class distribution for both labeled and unlabeled medical images. However, medical image data in reality is commonly imbalanced (i.e., multi-class label imbalance), which naturally yields blurry contours and usually incorrectly labels rare objects. Moreover, it remains unclear whether all negative samples are equally negative. In this work, we present ACTION, an Anatomical-aware ConTrastive dIstillatiON framework, for semi-supervised medical image segmentation. Specifically, we first develop an iterative contrastive distillation algorithm by softly labeling the negatives rather than binary supervision between positive and negative pairs. We also capture more semantically similar features from the randomly chosen negative set compared to the positives to enforce the diversity of the sampled data. Second, we raise a more important question: Can we really handle imbalanced samples to yield better performance? Hence, the key innovation in ACTION is to learn global semantic relationship across the entire dataset and local anatomical features among the neighbouring pixels with minimal additional memory footprint. During the training, we introduce anatomical contrast by actively sampling a sparse set of hard negative pixels, which can generate smoother segmentation boundaries and more accurate predictions. Extensive experiments across two benchmark datasets and different unlabeled settings show that ACTION significantly outperforms the current state-of-the-art semi-supervised methods.
{"title":"Bootstrapping Semi-supervised Medical Image Segmentation with Anatomical-aware Contrastive Distillation","authors":"Chenyu You, Weichen Dai, L. Staib, J. Duncan","doi":"10.48550/arXiv.2206.02307","DOIUrl":"https://doi.org/10.48550/arXiv.2206.02307","url":null,"abstract":"Contrastive learning has shown great promise over annotation scarcity problems in the context of medical image segmentation. Existing approaches typically assume a balanced class distribution for both labeled and unlabeled medical images. However, medical image data in reality is commonly imbalanced (i.e., multi-class label imbalance), which naturally yields blurry contours and usually incorrectly labels rare objects. Moreover, it remains unclear whether all negative samples are equally negative. In this work, we present ACTION, an Anatomical-aware ConTrastive dIstillatiON framework, for semi-supervised medical image segmentation. Specifically, we first develop an iterative contrastive distillation algorithm by softly labeling the negatives rather than binary supervision between positive and negative pairs. We also capture more semantically similar features from the randomly chosen negative set compared to the positives to enforce the diversity of the sampled data. Second, we raise a more important question: Can we really handle imbalanced samples to yield better performance? Hence, the key innovation in ACTION is to learn global semantic relationship across the entire dataset and local anatomical features among the neighbouring pixels with minimal additional memory footprint. During the training, we introduce anatomical contrast by actively sampling a sparse set of hard negative pixels, which can generate smoother segmentation boundaries and more accurate predictions. Extensive experiments across two benchmark datasets and different unlabeled settings show that ACTION significantly outperforms the current state-of-the-art semi-supervised methods.","PeriodicalId":73379,"journal":{"name":"Information processing in medical imaging : proceedings of the ... conference","volume":"13939 1","pages":"641-653"},"PeriodicalIF":0.0,"publicationDate":"2022-06-06","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"42815533","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2022-05-23DOI: 10.48550/arXiv.2205.11115
Manxi Lin, Zahra Bashir, M. Tolsgaard, A. Christensen, Aasa Feragen
Curvilinear structure segmentation is important in medical imaging, quantifying structures such as vessels, airways, neurons, or organ boundaries in 2D slices. Segmentation via pixel-wise classification often fails to capture the small and low-contrast curvilinear structures. Prior topological information is typically used to address this problem, often at an expensive computational cost, and sometimes requiring prior knowledge of the expected topology. We present DTU-Net, a data-driven approach to topology-preserving curvilinear structure segmentation. DTU-Net consists of two sequential, lightweight U-Nets, dedicated to texture and topology, respectively. While the texture net makes a coarse prediction using image texture information, the topology net learns topological information from the coarse prediction by employing a triplet loss trained to recognize false and missed splits in the structure. We conduct experiments on a challenging multi-class ultrasound scan segmentation dataset as well as a well-known retinal imaging dataset. Results show that our model outperforms existing approaches in both pixel-wise segmentation accuracy and topological continuity, with no need for prior topological knowledge.
{"title":"DTU-Net: Learning Topological Similarity for Curvilinear Structure Segmentation","authors":"Manxi Lin, Zahra Bashir, M. Tolsgaard, A. Christensen, Aasa Feragen","doi":"10.48550/arXiv.2205.11115","DOIUrl":"https://doi.org/10.48550/arXiv.2205.11115","url":null,"abstract":"Curvilinear structure segmentation is important in medical imaging, quantifying structures such as vessels, airways, neurons, or organ boundaries in 2D slices. Segmentation via pixel-wise classification often fails to capture the small and low-contrast curvilinear structures. Prior topological information is typically used to address this problem, often at an expensive computational cost, and sometimes requiring prior knowledge of the expected topology. We present DTU-Net, a data-driven approach to topology-preserving curvilinear structure segmentation. DTU-Net consists of two sequential, lightweight U-Nets, dedicated to texture and topology, respectively. While the texture net makes a coarse prediction using image texture information, the topology net learns topological information from the coarse prediction by employing a triplet loss trained to recognize false and missed splits in the structure. We conduct experiments on a challenging multi-class ultrasound scan segmentation dataset as well as a well-known retinal imaging dataset. Results show that our model outperforms existing approaches in both pixel-wise segmentation accuracy and topological continuity, with no need for prior topological knowledge.","PeriodicalId":73379,"journal":{"name":"Information processing in medical imaging : proceedings of the ... conference","volume":"318 1","pages":"654-666"},"PeriodicalIF":0.0,"publicationDate":"2022-05-23","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"76439568","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2022-05-17DOI: 10.48550/arXiv.2205.08209
F. Kofler, Suprosanna Shit, I. Ezhov, L. Fidon, Rami Al-Maskari, Hongwei Li, H. Bhatia, T. Loehr, M. Piraud, Ali Erturk, J. Kirschke, J. Peeken, Tom Kamiel Magda Vercauteren, C. Zimmer, B. Wiestler, Bjoern H Menze
Deep convolutional neural networks (CNN) have proven to be remarkably effective in semantic segmentation tasks. Most popular loss functions were introduced targeting improved volumetric scores, such as the Dice coefficient (DSC). By design, DSC can tackle class imbalance, however, it does not recognize instance imbalance within a class. As a result, a large foreground instance can dominate minor instances and still produce a satisfactory DSC. Nevertheless, detecting tiny instances is crucial for many applications, such as disease monitoring. For example, it is imperative to locate and surveil small-scale lesions in the follow-up of multiple sclerosis patients. We propose a novel family of loss functions, emph{blob loss}, primarily aimed at maximizing instance-level detection metrics, such as F1 score and sensitivity. emph{Blob loss} is designed for semantic segmentation problems where detecting multiple instances matters. We extensively evaluate a DSC-based emph{blob loss} in five complex 3D semantic segmentation tasks featuring pronounced instance heterogeneity in terms of texture and morphology. Compared to soft Dice loss, we achieve 5% improvement for MS lesions, 3% improvement for liver tumor, and an average 2% improvement for microscopy segmentation tasks considering F1 score.
深度卷积神经网络(CNN)在语义分割任务中已经被证明是非常有效的。最流行的损失函数是针对改进的体积分数，如Dice系数(DSC)。通过设计，DSC可以处理类的不平衡，但是，它不能识别类中的实例不平衡。因此，大型前台实例可以支配较小的实例，并且仍然产生令人满意的DSC。然而，检测微小实例对于许多应用程序(如疾病监测)至关重要。例如，在多发性硬化症患者的随访中，定位和监测小范围病变是必不可少的。我们提出了一种新的损失函数，emph{blob损失}，主要目的是最大化实例级检测指标，如F1分数和灵敏度。emph{Blob损失}是为语义分割问题而设计的，其中检测多个实例很重要。我们在五个复杂的3D语义分割任务中广泛评估了基于dsc的emph{blob损失}，这些任务在纹理和形态方面具有明显的实例异质性。与软骰子损失相比，我们达到了5% improvement for MS lesions, 3% improvement for liver tumor, and an average 2% improvement for microscopy segmentation tasks considering F1 score.
{"title":"Blob Loss: Instance Imbalance Aware Loss Functions for Semantic Segmentation","authors":"F. Kofler, Suprosanna Shit, I. Ezhov, L. Fidon, Rami Al-Maskari, Hongwei Li, H. Bhatia, T. Loehr, M. Piraud, Ali Erturk, J. Kirschke, J. Peeken, Tom Kamiel Magda Vercauteren, C. Zimmer, B. Wiestler, Bjoern H Menze","doi":"10.48550/arXiv.2205.08209","DOIUrl":"https://doi.org/10.48550/arXiv.2205.08209","url":null,"abstract":"Deep convolutional neural networks (CNN) have proven to be remarkably effective in semantic segmentation tasks. Most popular loss functions were introduced targeting improved volumetric scores, such as the Dice coefficient (DSC). By design, DSC can tackle class imbalance, however, it does not recognize instance imbalance within a class. As a result, a large foreground instance can dominate minor instances and still produce a satisfactory DSC. Nevertheless, detecting tiny instances is crucial for many applications, such as disease monitoring. For example, it is imperative to locate and surveil small-scale lesions in the follow-up of multiple sclerosis patients. We propose a novel family of loss functions, emph{blob loss}, primarily aimed at maximizing instance-level detection metrics, such as F1 score and sensitivity. emph{Blob loss} is designed for semantic segmentation problems where detecting multiple instances matters. We extensively evaluate a DSC-based emph{blob loss} in five complex 3D semantic segmentation tasks featuring pronounced instance heterogeneity in terms of texture and morphology. Compared to soft Dice loss, we achieve 5% improvement for MS lesions, 3% improvement for liver tumor, and an average 2% improvement for microscopy segmentation tasks considering F1 score.","PeriodicalId":73379,"journal":{"name":"Information processing in medical imaging : proceedings of the ... conference","volume":"85 1","pages":"755-767"},"PeriodicalIF":0.0,"publicationDate":"2022-05-17","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"83889020","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2022-03-06DOI: 10.1007/978-3-031-34048-2_23
Haocheng Dai, M. Bauer, P. Fletcher, S. Joshi
{"title":"Modeling the Shape of the Brain Connectome via Deep Neural Networks","authors":"Haocheng Dai, M. Bauer, P. Fletcher, S. Joshi","doi":"10.1007/978-3-031-34048-2_23","DOIUrl":"https://doi.org/10.1007/978-3-031-34048-2_23","url":null,"abstract":"","PeriodicalId":73379,"journal":{"name":"Information processing in medical imaging : proceedings of the ... conference","volume":"14 1","pages":"291-302"},"PeriodicalIF":0.0,"publicationDate":"2022-03-06","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"90181431","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
{"title":"Segmentation with Multiple Acceptable Annotations: A Case Study of Myocardial Segmentation in Contrast Echocardiography","authors":"Dewen Zeng, Mingqi Li, Yukun Ding, Xiaowei Xu, Qiu Xie, Ruixue Xu, Hongwen Fei, Meiping Huang, Zhuang Jian, Yiyu Shi","doi":"10.1007/978-3-030-78191-0_37","DOIUrl":"https://doi.org/10.1007/978-3-030-78191-0_37","url":null,"abstract":"","PeriodicalId":73379,"journal":{"name":"Information processing in medical imaging : proceedings of the ... conference","volume":"91 1","pages":"478-491"},"PeriodicalIF":0.0,"publicationDate":"2021-06-29","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"78329412","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2021-06-28DOI: 10.1007/978-3-030-78191-0_24
Jose J. Bouza, Chun-Hao Yang, D. Vaillancourt, B. Vemuri
{"title":"A Higher Order Manifold-Valued Convolutional Neural Network with Applications to Diffusion MRI Processing","authors":"Jose J. Bouza, Chun-Hao Yang, D. Vaillancourt, B. Vemuri","doi":"10.1007/978-3-030-78191-0_24","DOIUrl":"https://doi.org/10.1007/978-3-030-78191-0_24","url":null,"abstract":"","PeriodicalId":73379,"journal":{"name":"Information processing in medical imaging : proceedings of the ... conference","volume":"23 1","pages":"304-317"},"PeriodicalIF":0.0,"publicationDate":"2021-06-28","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"72852263","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2021-06-28DOI: 10.1007/978-3-030-78191-0_18
Jiazhou Chen, Defu Yang, Hongmin Cai, M. Styner, Guorong Wu
{"title":"Discovering Spreading Pathways of Neuropathological Events in Alzheimer's Disease Using Harmonic Wavelets","authors":"Jiazhou Chen, Defu Yang, Hongmin Cai, M. Styner, Guorong Wu","doi":"10.1007/978-3-030-78191-0_18","DOIUrl":"https://doi.org/10.1007/978-3-030-78191-0_18","url":null,"abstract":"","PeriodicalId":73379,"journal":{"name":"Information processing in medical imaging : proceedings of the ... conference","volume":"25 1 1","pages":"228-240"},"PeriodicalIF":0.0,"publicationDate":"2021-06-28","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"88070000","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2021-06-07DOI: 10.1007/978-3-030-78191-0_50
Matthias Perkonigg, J. Hofmanninger, G. Langs
{"title":"Continual Active Learning for Efficient Adaptation of Machine Learning Models to Changing Image Acquisition","authors":"Matthias Perkonigg, J. Hofmanninger, G. Langs","doi":"10.1007/978-3-030-78191-0_50","DOIUrl":"https://doi.org/10.1007/978-3-030-78191-0_50","url":null,"abstract":"","PeriodicalId":73379,"journal":{"name":"Information processing in medical imaging : proceedings of the ... conference","volume":"40 1","pages":"649-660"},"PeriodicalIF":0.0,"publicationDate":"2021-06-07","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"76468866","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
京公网安备 11010802042870号
京ICP备2023020795号-1
扫码关注我们
求助内容：
应助结果提醒方式：