Proceedings. IEEE International Symposium on Biomedical Imaging最新文献

This paper presents a multimodal deep learning framework that utilizes advanced image techniques to improve the performance of clinical analysis heavily dependent on routinely acquired standard images. More specifically, we develop a joint learning network that for the first time leverages the accuracy and reproducibility of myocardial strains obtained from Displacement Encoding with Stimulated Echo (DENSE) to guide the analysis of cine cardiac magnetic resonance (CMR) imaging in late mechanical activation (LMA) detection. An image registration network is utilized to acquire the knowledge of cardiac motions, an important feature estimator of strain values, from standard cine CMRs. Our framework consists of two major components: (i) a DENSE-supervised strain network leveraging latent motion features learned from a registration network to predict myocardial strains; and (ii) a LMA network taking advantage of the predicted strain for effective LMA detection. Experimental results show that our proposed work substantially improves the performance of strain analysis and LMA detection from cine CMR images, aligning more closely with the achievements of DENSE.

Retinotopic mapping aims to uncover the relationship between visual stimuli on the retina and neural responses on the visual cortical surface. This study advances retinotopic mapping by applying diffeomorphic registration to the 3T NYU retinotopy dataset, encompassing analyze-PRF and mrVista data. Diffeomorphic Registration for Retinotopic Maps (DRRM) quantifies the diffeomorphic condition, ensuring accurate alignment of retinotopic maps without topological violations. Leveraging the Beltrami coefficient and topological condition, DRRM significantly enhances retinotopic map accuracy. Evaluation against existing methods demonstrates DRRM's superiority on various datasets, including 3T and 7T retinotopy data. The application of diffeomorphic registration improves the interpretability of low-quality retinotopic maps, holding promise for clinical applications.

Mild cognitive impairment (MCI) is recognized as a precursor to Alzheimer's disease (AD), a progressive and irreversible neurodegenerative disorder of the brain. The neurodegeneration of brain connectivity networks plays a pivotal role in the development and progression of MCI. Traditionally, brain networks are generated using coarse-grained brain regions, where the regions serve as nodes and their functional or structural connections are used as edges. Recently, a novel finer scale brain folding patterns named 3-hinge gyrus (3HG) was identified, which is defined as the conjunctions coming from three directions on gyral crests. 3HGs have been shown playing an important role in brain network and can serve as hubs. In this study, our objective is to construct a novel 3HG-based finer-scale brain connectome and comprehensively compare its performance with traditional region-based connectome in predicting MCI against Normal Controls (NC). The results of extensive experiments demonstrate the superior performance of 3HG-based brain connectome, shedding light on the potential of 3HG-based connectomes in capturing intricate neurodegenerative patterns associated with MCI and AD.

Recently, a novel cortical folding pattern known as the 3-hinge gyrus (3HG) has been identified. 3HGs are defined as the convergence of the gyri coming from three distinct directions on gyral crests. In contrast to cortical regions, 3HGs are defined at a finer scale and they widely exist across different individuals, representing both commonalities and individualities of cortical folding patterns. It is important to note that 3HGs are identified in individual spaces, lacking natural cross-subject correspondences. To address this issue, we have developed a learning-based method to encode anatomical features of 3HGs into a set of embedding vectors that can be compared across individuals. However, this method solely relies on anatomical features and can be suboptimal because it does not consider the related structural connectivity patterns, as many 3HGs have multiple potential matches using anatomical properties only. In this study, we leverage the multimodal imaging data (T1 MRI and DTI) which are complementary to each other in representing 3HGs, to enhance the precision when identifying one-to-one correspondence for 3HGs. Through extensive experiments, we have demonstrated the effectiveness of our approach in mitigating the one-to-many match issue associated with 3HGs, significantly improving the accuracy of 3HG correspondences. This accomplishment holds considerable implications for group-level analyses based on 3HGs and contributes to the broader utilization of 3HGs in brain studies.

Arterial spin-labeled (ASL) perfusion MRI remains the only non-invasive, radiation-free method for quantifying regional tissue perfusion. ASL MRI computes perfusion signals from the difference of the spin-labeled images and spin-untagged control images. Limited by the T1 decay of the labeled arterial blood, ASL MRI signal is subject to a low signal-to-noise ratio. This issue is particularly vexing due to the absence of ground truth and the difficulty in preserving image textures amidst substantial noise reduction efforts. One major avenue for tackling this challenge involves leveraging the sparsity of image signals, a technique widely employed in unsupervised image denoising. Compared to global models operating at the slice level, enhanced local sparse models not only improve the separation of signal from noise but also preserves local structures more effectively. This paper introduces a joint data selection strategy tailored for ASL denoising, which capitalizes on the strong correlation between paired label and control (L/C) images to identify and assemble highly correlated content, forming potentially sparse matrices. The application of sparsity regularization to these matrices is inherently more adaptive to local structures. Crucially, the proposed method does not rely on any ground-truth training data. In real-world testing with an ASL MRI dataset, the proposed approach remarkably enhances the quality of ASL perfusion maps, utilizing only a single pair of L/C images, and outperforms the conventional pipeline that necessitates multiple L/C pairs.

Functional MRI (fMRI) is an important tool for non-invasive studies of brain function. Over the past decade, multi-echo fMRI methods that sample multiple echo times has become popular with potential to improve quantification. While these acquisitions are typically performed with Cartesian trajectories, non-Cartesian trajectories, in particular spiral acquisitions, hold promise for denser sampling of echo times. However, such acquisitions require very high acceleration rates for sufficient spatiotemporal resolutions. In this work, we propose to use a physics-driven deep learning (PD-DL) reconstruction to accelerate multi-echo spiral fMRI by 10-fold. We modify a self-supervised learning algorithm for optimized training with non-Cartesian trajectories and use it to train the PD-DL network. Results show that the proposed self-supervised PD-DL reconstruction achieves high spatio-temporal resolution with meaningful BOLD analysis.

Segmentation of brain tumors in multi-parametric magnetic resonance imaging facilitates quantitative analysis crucial for clinical trials and personalized patient care. This significantly influences clinical decision-making, encompassing diagnosis and prognosis and enhancing patient outcomes. The brain tumor segmentation (BraTS) challenge, in its 2023 edition, extended to a cluster of competitions incorporating multiple tumor types. Now, in conjunction with IEEE ISBI 2024, BraTS organizes its Generalizability Across Tumors (BraTS-GoAT) challenge. In this paper, we introduce a deep-learning-based ensemble strategy involving three state-of-the-art segmentation models. Furthermore, we also introduce a novel adaptive post-processing method, based on a cross-validated tumor-specific threshold search, designed to output enhanced accurate segmentations, ensuring generalizability across various tumor types. The evaluation of our proposed method on validation cases resulted in lesion-wise Dice scores of 0.842, 0.854, 0.872 and lesion-wise 95th-percentile Hausdorff Distance scores of 29.46, 24.67, 25.22 for the enhancing tumor, tumor core, and whole tumor, respectively.

Spatially aligning two computed tomography (CT) scans of the lung using automated image registration techniques is a challenging task due to the deformable nature of the lung. However, existing deep-learning-based lung CT registration models are not trained with explicit anatomical knowledge. We propose the deformable anatomy-aware registration toolkit (DART), a masked autoencoder (MAE)-based approach, to improve the keypoint-supervised registration of lung CTs. Our method incorporates features from multiple decoders of networks trained to segment anatomical structures, including the lung, ribs, vertebrae, lobes, vessels, and airways, to ensure that the MAE learns relevant features corresponding to the anatomy of the lung. The pretrained weights of the transformer encoder and patch embeddings are then used as the initialization for the training of downstream registration. We compare DART to existing state-of-the-art registration models. Our experiments show that DART outperforms the baseline models (Voxelmorph, ViT-V-Net, and MAE-TransRNet) in terms of target registration error of both corrField-generated keypoints with 17%, 13%, and 9% relative improvement, respectively, and bounding box centers of nodules with 27%, 10%, and 4% relative improvement, respectively. Our implementation is available at https://github.com/yunzhengzhu/DART.

The complex etiology of various neurodegenerative diseases and psychiatric disorders, especially at the individual level, has posed unmatched challenges to the advancement of personalized medicine. Recent technical advancements in functional magnetic resonance imaging has enabled researchers to map brain large-scale connectivity at an unprecedented level of subject precision. Nonetheless, along with the early dawn of promises in personalized medicine using various neuroimaging modalities rose the challenge of clinical utility of brain connectomics (e.g., functional connectomes). Besides many established challenges of functional connectome utility such as edge reliability, there exists an easily overlooked challenge that does not get the same level of attention: computationality of functional connectome. To improve clinical utility of functional connectomics, we propose a random projection method that would preserve a practically similar level of subject identifiability while sampling and retaining only a proportion of functional edges in subjects' functional connectome. Our work pave a way towards computational improvements, hence clinical utility, of functional connectomes while not compromising the integrity of biomarkers learnt from whole-brain large-scale functional connectivity imaging modality.

Hippocampal atrophy in Alzheimer's disease (AD) is asymmetric and spatially inhomogeneous. While extensive work has been done on volume and shape analysis of atrophy of the hippocampus in AD, less attention has been given to hippocampal asymmetry specifically. Previous studies of hippocampal asymmetry are limited to global volume or shape measures, which don't localize shape asymmetry at the point level. In this paper, we propose to quantify localized shape asymmetry by optimizing point correspondences between left and right hippocampi within a subject, while simultaneously favoring a compact statistical shape model of the entire sample. To account for related variables that have an impact on AD and healthy subject differences, we build linear models with other confounding factors. Our results on the OASIS3 dataset demonstrate that compared to volumetric information, shape asymmetry reveals fine-grained, localized differences that inform us about the hippocampal regions of most significant shape asymmetry in AD patients.