检索结果-内蒙古大学图书馆

DEALing with Image Reconstruction: Deep Attentive Least Squares

学校读者我要写书评

暂无评论

arXiv 2025年

作者： Pourya, Mehrsa Kobler, Erich Unser, Michael Neumayer, Sebastian Biomedical Imaging Group EPFL Lausanne Switzerland Institute for Machine Learning LIT AI lab Institute for Virtual Morphology Johannes Kepler University Linz Austria Faculty of Mathematics TU Chemnitz Germany

State-of-the-art image reconstruction often relies on complex, highly parameterized deep architectures. We propose an alternative: a data-driven reconstruction method inspired by the classic Tikhonov regularization. Our approach iteratively refines intermediate reconstructions by solving a sequence of quadratic problems. These updates have two key components: (i) learned filters to extract salient image features, and (ii) an attention mechanism that locally adjusts the penalty of filter responses. Our method achieves performance on par with leading plug-and-play and learned regularizer approaches while offering interpretability, robustness, and convergent behavior. In effect, we bridge traditional regularization and deep learning with a principled reconstruction approach. © 2025, CC BY.

关键词： Image reconstruction

Fair and Private CT Contrast Agent Detection 2nd

学校读者我要写书评

暂无评论

Fair and Private CT Contrast Agent Detection

2nd International Workshop on Fairness of AI in Medical imaging, FAIMI 2024, and 3rd International Workshop on Ethical and Philosophical Issues in Medical imaging, EPIMI 2024, Held in Conjunction with the International Conference on Medical Image Computing and Computer Assisted Interventions, MICCAI 2024

作者： Kaess, Philipp Ziller, Alexander Mantz, Lea Rueckert, Daniel Fintelmann, Florian J. Kaissis, Georgios AI in Healthcare and Medicine Technical University of Munich Munich Germany Department of Radiology Massachusetts General Hospital Boston United States Department of Diagnostic and Interventional Radiology University Medical Center of the Johannes Gutenberg University Mainz Mainz Germany Department of Computing Imperial College London London United Kingdom Institute of Machine Learning in Biomedical Imaging Helmholtz Munich Germany

ISBN: (纸本)9783031727863

Intravenous (IV) contrast agents are an established medical tool to enhance the visibility of certain structures. However, their application substantially changes the appearance of Computed Tomography (CT) images, which - if unknown - can significantly deteriorate the diagnostic performance of neural networks. Artificial Intelligence (AI) can help to detect IV contrast, reducing the need for labour-intensive and error-prone manual labeling. However, we demonstrate that automated contrast detection can lead to discrimination against demographic subgroups. Moreover, it has been shown repeatedly that AI models can leak private training data. In this work, we analyse the fairness of conventional and privacy-preserving AI models during the detection of IV contrast on CT images. Specifically, we present models which are substantially fairer compared to a previously published baseline. For better comparability, we extend existing metrics to quantify the fairness of a model on a protected attribute in a single value. We provide a model, fulfilling a strict Differential Privacy protection of (Ε,δ)=(8,2.8·10-3), which with an accuracy of 97.42% performs 5%-points better than the baseline. Additionally, while confirming prior works, that strict privacy preservation increases the discrimination against underrepresented subgroups, the proposed model is fairer than the baseline over all metrics considering race and sex as protected attributes, which extends to age for a more relaxed privacy guarantee. © The Author(s), under exclusive license to Springer Nature Switzerland AG 2025.

关键词： Differential privacy

Cross-Domain and Cross-Dimension learning for Image-to-Graph Transformers

学校读者我要写书评

暂无评论

Cross-Domain and Cross-Dimension Learning for Image-to-Graph...

2025 IEEE/CVF Winter Conference on Applications of Computer Vision, WACV 2025

作者： Berger, Alexander H. Lux, Laurin Shit, Suprosanna Ezhov, Ivan Kaissis, Georgios Menten, Martin J. Rueckert, Daniel Paetzold, Johannes C. School of Computation Information and Technology Technical University of Munich Germany Department of Computing Imperial College London United Kingdom Munich Germany Institute for Machine Learning in Biomedical Imaging Helmholtz Munich Germany Department of Quantitative Biomedicine University of Zurich Switzerland Weill Cornell Medicine Cornell University New York City United States

ISBN: (纸本)9798331510831

Direct image-to-graph transformation is a challenging task that involves solving object detection and relationship prediction in a single model. Due to this task's complexity, large training datasets are rare in many domains, making the training of deep-learning methods challenging. This data sparsity necessitates transfer learning strategies akin to the state-of-the-art in general computer vision. In this work, we introduce a set of methods enabling cross-domain and cross-dimension learning for image-to-graph trans-formers. We propose (1) a regularized edge sampling loss to effectively learn object relations in multiple domains with different numbers of edges, (2) a domain adaptation frame-work for image-to-graph transformers aligning image- and graph-level features from different domains, and (3) a projection function that allows using 2D data for training 3D transformers. We demonstrate our method's utility in cross-domain and cross-dimension experiments, where we utilize labeled data from 2D road networks for simultaneous learning in vastly different target domains. Our method consistently outperforms standard transfer learning and self-supervised pretraining on challenging benchmarks, such as retinal or whole-brain vessel graph extraction.11Code: ***/AlexanderHBerger/cross-dim_i2g © 2025 IEEE.

关键词： Self-supervised learning

TGV: Tabular Data-Guided learning of Visual Cardiac Representations

学校读者我要写书评

暂无评论

arXiv 2025年

作者： Hasny, Marta Folco, Maxime Di Bressem, Keno Schnabel, Julia School of Computation Information and Technology Technical University of Munich Germany Institute of Machine Learning for Biomedical Imaging Helmholtz Munich Germany German Heart Center Munich Technical University of Munich Germany Klinikum rechts der Isar Technical University of Munich Germany School of Biomedical Engineering & Imaging Sciences King’s College London United Kingdom

Contrastive learning methods in computer vision typically rely on different views of the same image to form pairs. However, in medical imaging, we often seek to compare entire patients with different phenotypes rather than just multiple augmentations of one scan. We propose harnessing clinically relevant tabular data to identify distinct patient phenotypes and form more meaningful pairs in a contrastive learning framework. Our method uses tabular attributes to guide the training of visual representations, without requiring a joint embedding space. We demonstrate its strength using short-axis cardiac MR images and clinical attributes from the UK Biobank, where tabular data helps to more effectively distinguish between patient subgroups. Evaluation on downstream tasks, including fine-tuning and zero-shot prediction of cardiovascular artery diseases and cardiac phenotypes, shows that incorporating tabular data yields stronger visual representations than conventional methods that rely solely on image augmentations or combined image-tabular embeddings. Furthermore, we demonstrate that image encoders trained with tabular guidance are capable of embedding demographic information in their representations, allowing them to use insights from tabular data for unimodal predictions, making them well-suited to real-world medical settings where extensive clinical annotations may not be routinely available at inference time. The code will be available on GitHub. © 2025, CC BY-NC-ND.

关键词： Contrastive learning

Cross-Domain and Cross-Dimension learning for Image-to-Graph Transformers

学校读者我要写书评

暂无评论

Cross-Domain and Cross-Dimension Learning for Image-to-Graph...

IEEE Workshop on Applications of Computer Vision (WACV)

作者： Alexander H. Berger Laurin Lux Suprosanna Shit Ivan Ezhov Georgios Kaissis Martin J. Menten Daniel Rueckert Johannes C. Paetzold School of Computation Information and Technology Technical University of Munich Germany Munich Center for Machine Learning (MCML) Munich Germany Department of Quantitative Biomedicine University of Zurich Switzerland Institute for Machine Learning in Biomedical Imaging Helmholtz Munich Germany Department of Computing Imperial College London UK Weill Cornell Medicine Cornell University New York City USA

ISBN: (数字)9798331510831

ISBN: (纸本)9798331510848

关键词： Training Computer vision Three-dimensional displays Image edge detection Roads Transfer learning Predictive models Transformers Retina Standards

High-dimensional multimodal uncertainty estimation by manifold alignment: Application to 3D right ventricular strain computations

学校读者我要写书评

暂无评论

arXiv 2025年

作者： Folco, Maxime Di Bernardino, Gabriel Clarysse, Patrick Duchateau, Nicolas Univ Lyon Université Claude Bernard Lyon 1 INSA-Lyon CNRS Inserm CREATIS UMR 5220 U1294 LyonF-69621 France Institute of Machine Learning in Biomedical Imaging Helmholtz Center Munich Germany DTIC Universitat Pompeu Fabra Barcelona Spain France

Confidence in the results is a key ingredient to improve the adoption of machine learning methods by clinicians. Uncertainties on the results have been considered in the literature, but mostly those originating from the learning and processing methods. Uncertainty on the data is hardly challenged, as a single sample is often considered representative enough of each subject included in the analysis. In this paper, we propose a representation learning strategy to estimate local uncertainties on a physiological descriptor (here, myocardial deformation) previously obtained from medical images by different definitions or computations. We first use manifold alignment to match the latent representations associated to different high-dimensional input descriptors. Then, we formulate plausible distributions of latent uncertainties, and finally exploit them to reconstruct uncertainties on the input high-dimensional descriptors. We demonstrate its relevance for the quantification of myocardial deformation (strain) from 3D echocardiographic image sequences of the right ventricle, for which a lack of consensus exists in its definition and which directional component to use. We used a database of 100 control subjects with right ventricle overload, for which different types of strain are available at each point of the right ventricle endocardial surface mesh. Our approach quantifies local uncertainties on myocardial deformation from different descriptors defining this physiological concept. Such uncertainties cannot be directly estimated by local statistics on such descriptors, potentially of heterogeneous types. Beyond this controlled illustrative application, our methodology has the potential to be generalized to many other population analyses considering heterogeneous high-dimensional descriptors. © 2025, CC BY-NC-ND.

关键词： Echocardiography

PISCO: Self-Supervised k-Space Regularization for Improved Neural Implicit k-Space Representations of Dynamic MRI

学校读者我要写书评

暂无评论

arXiv 2025年

作者： Spieker, Veronika Eichhorn, Hannah Huang, Wenqi Stelter, Jonathan K. Catalan, Tabita Braren, Rickmer F. Rueckert, Daniel Costabal, Francisco Sahli Hammernik, Kerstin Karampinos, Dimitrios C. Prieto, Claudia Schnabel, Julia A. Institute of Machine Learning for Biomedical Imaging Helmholtz Munich Neuherberg Germany School of Computation Information and Technology Technical University of Munich Germany Millenium Institute for Intelligent Healthcare Engineering Santiago Chile School of Medicine and Health Klinikum rechts der Isar Technical University of Munich Munich Germany Department of Computing Imperial College London London United Kingdom School of Engineering Pontificia Universidad Católica de Chile Santiago Chile School of Biomedical Imaging and Imaging Sciences King’s College London London United Kingdom

Neural implicit k-space representations (NIK) have shown promising results for dynamic magnetic resonance imaging (MRI) at high temporal resolutions. Yet, reducing acquisition time, and thereby available training data, results in severe performance drops due to overfitting. To address this, we introduce a novel self-supervised k-space loss function LPISCO, applicable for regularization of NIK-based reconstructions. The proposed loss function is based on the concept of parallel imaging-inspired self-consistency (PISCO), enforcing a consistent global k-space neighborhood relationship without requiring additional data. Quantitative and qualitative evaluations on static and dynamic MR reconstructions show that integrating PISCO significantly improves NIK representations. Particularly for high acceleration factors (R≥54), NIK with PISCO achieves superior spatiotemporal reconstruction quality compared to state-of-the-art methods. Furthermore, an extensive analysis of the loss assumptions and stability shows PISCO’s potential as versatile self-supervised k-space loss function for further applications and architectures. Code is available at: https://***/compai-lab/ 2025-pisco-spieker Copyright © 2025, The Authors. All rights reserved.

关键词： Dynamic contrast enhanced MRI

Motion-Robust T∗2 Quantification from Gradient Echo MRI with Physics-Informed Deep learning

学校读者我要写书评

暂无评论

arXiv 2025年

作者： Eichhorn, Hannah Spieker, Veronika Hammernik, Kerstin Saks, Elisa Felsner, Lina Weiss, Kilian Preibisch, Christine Schnabel, Julia A. Institute of Machine Learning in Biomedical Imaging Helmholtz Munich Germany School of Computation Information & Technology Technical University of Munich Germany School of Medicine & Health Institute for Diagnostic and Interventional Neuroradiology Technical University of Munich Germany School of Medicine & Health TUM-Neuroimaging Center Technical University of Munich Germany Philips GmbH Market DACH Germany School of Medicine & Health Clinic of Neurology Technical University of Munich Germany School of Biomedical Engineering & Imaging Sciences King’s College London United Kingdom

Purpose: T∗2 quantification from gradient echo magnetic resonance imaging is particularly affected by subject motion due to the high sensitivity to magnetic field inhomogeneities, which are influenced by motion and might cause signal loss. Thus, motion correction is crucial to obtain high-quality T∗2 maps. Methods: We extend our previously introduced learning-based physics-informed motion correction method, PHIMO, by utilizing acquisition knowledge to enhance the reconstruction performance for challenging motion patterns and increase PHIMO’s robustness to varying strengths of magnetic field inhomogeneities across the brain. We perform comprehensive evaluations regarding motion detection accuracy and image quality for data with simulated and real motion. Results: Our extended version of PHIMO outperforms the learning-based baseline methods both qualitatively and quantitatively with respect to line detection and image quality. Moreover, PHIMO performs on-par with a conventional state-of-the-art motion correction method for T∗2 quantification from gradient echo MRI, which relies on redundant data acquisition. Conclusion: PHIMO’s competitive motion correction performance, combined with a reduction in acquisition time by over 40% compared to the state-of-the-art method, make it a promising solution for motion-robust T∗2 quantification in research settings and clinical routine. Copyright © 2025, The Authors. All rights reserved.

关键词： Magnetic fields

AI-DRIVEN AUTOMATED TOOL FOR ABDOMINAL CT BODY COMPOSITION ANALYSIS IN GASTROINTESTINAL CANCER MANAGEMENT

学校读者我要写书评

暂无评论

arXiv 2025年

作者： Nan, Xinyu He, Meng Chen, Zifan Dong, Bin Tang, Lei Zhang, Li Center for Data Science Peking University China Department of Radiology Key Laboratory of Carcinogenesis and Translational Research Ministry of Education Peking University Cancer Hospital and Institute Beijing China Peking University Beijing China Center for Machine Learning Research Peking University Beijing China National Biomedical Imaging Center Peking University Beijing China

The incidence of gastrointestinal cancers remains significantly high, particularly in China, emphasizing the importance of accurate prognostic assessments and effective treatment strategies. Research shows a strong correlation between abdominal muscle and fat tissue composition and patient outcomes. However, existing manual methods for analyzing abdominal tissue composition are time-consuming and costly, limiting clinical research scalability. To address these challenges, we developed an AI-driven tool for automated analysis of abdominal CT scans to effectively identify and segment muscle, subcutaneous fat, and visceral fat. Our tool integrates a multi-view localization model and a high-precision 2D nnUNet-based segmentation model, demonstrating a localization accuracy of 90% and a Dice Score Coefficient of 0.967 for segmentation. Furthermore, it features an interactive interface that allows clinicians to refine the segmentation results, ensuring high-quality outcomes effectively. Our tool offers a standardized method for effectively extracting critical abdominal tissues, potentially enhancing the management and treatment for gastrointestinal cancers. The code is available at https://***/NanXinyu/***. Copyright © 2025, The Authors. All rights reserved.

关键词： Diseases