Article
Computer Science, Interdisciplinary Applications
Xiaodan Wei, Qinghao Liu, Min Liu, Yaonan Wang, Erik Meijering
Summary: In this paper, we propose a two-stage deep neural network for fast and accurate soma detection in large-scale and high-resolution whole mouse brain images. A lightweight Multi-level Cross Classification Network (MCC-Net) is first used to filter out images without somas and generate coarse candidate images, followed by a Scale Fusion Segmentation Network (SFS-Net) to accurately segment soma regions. Experimental results demonstrate excellent performance of the proposed method and a public dataset named WBMSD is established for further research on soma detection.
IEEE TRANSACTIONS ON MEDICAL IMAGING
(2023)
Article
Computer Science, Interdisciplinary Applications
Chao Liu, Deli Wang, Han Zhang, Wei Wu, Wenzhi Sun, Ting Zhao, Nenggan Zheng
Summary: Reconstructing neuron morphologies from fluorescence microscope images is crucial for neuroscience studies. This study proposes a strategy of using two-stage generative models to simulate training data with voxel-level labels, resulting in realistic 3D images with underlying voxel labels. The results show that networks trained on synthetic data outperform those trained on manually labeled data in segmentation performance.
IEEE TRANSACTIONS ON MEDICAL IMAGING
(2022)
Article
Computer Science, Interdisciplinary Applications
Bo Yang, Min Liu, Yaonan Wang, Kang Zhang, Erik Meijering
Summary: This paper presents a 3D neuron segmentation network called SGSNet that enhances weak neuronal structures and removes background noises. The network utilizes two decoding paths, one for acquiring segmentation maps and the other for detecting neuronal structures. A structure attention module is designed to integrate features and provide contextual guidance to improve segmentation performance.
IEEE TRANSACTIONS ON MEDICAL IMAGING
(2022)
Article
Computer Science, Interdisciplinary Applications
Weixun Chen, Min Liu, Hao Du, Miroslav Radojevic, Yaonan Wang, Erik Meijering
Summary: This paper proposes a novel method called SPE-DNR that combines spherical-patches extraction and deep-learning for neuron reconstruction. Experimental results demonstrate that the method is competitive and robust.
IEEE TRANSACTIONS ON MEDICAL IMAGING
(2022)
Article
Biochemical Research Methods
Qiufu Li, Linlin Shen
Summary: The paper introduces a 3D wavelet and deep learning-based method for neuron segmentation, utilizing 3D WaveUNet to process neuronal cubes and improve performance in noisy neuronal images. The integrated 3D wavelets efficiently assist in 3D neuron segmentation and reconstruction.
Article
Computer Science, Software Engineering
Parmida Ghahremani, Saeed Boorboor, Pooya Mirhosseini, Chetan Gudisagar, Mala Ananth, David Talmage, Lorna W. Role, Arie E. Kaufman
Summary: NeuroConstruct is a novel end-to-end application for the segmentation, registration, and visualization of brain volumes imaged using wide-field microscopy. It offers various annotation helper functions for precise and effective neurite annotation, as well as automatic neurite segmentation using convolutional neural networks. The application also includes tools for visualizing neurites and aligning serially sectioned samples.
IEEE TRANSACTIONS ON VISUALIZATION AND COMPUTER GRAPHICS
(2022)
Article
Computer Science, Interdisciplinary Applications
Kisuk Lee, Ran Lu, Kyle Luther, H. Sebastian Seung
Summary: This study demonstrates that dense voxel embeddings learned through deep metric learning can accurately segment neurons from 3D electron microscopy images. A metric graph is constructed from the dense voxel embeddings generated by a convolutional network, and partitioning the graph with long-range edges as repulsive constraints results in precise initial segmentation, particularly useful for very thin objects. The convolutional embedding network is reused without modification to aggregate systematic splits caused by complex self-contact motifs, achieving state-of-the-art accuracy in reconstructing 3D neurons from brain images acquired through serial section electron microscopy. The proposed object-centered representation may have broader applications in automated neural circuit reconstruction.
IEEE TRANSACTIONS ON MEDICAL IMAGING
(2021)
Article
Computer Science, Interdisciplinary Applications
Tianfang Zhu, Gang Yao, Dongli Hu, Chuangchuang Xie, Pengcheng Li, Xiaoquan Yang, Hui Gong, Qingming Luo, Anan Li
Summary: This study proposes MorphoGNN, a single neuron morphological embedding based on a graph neural network. By considering the point-level structure information of reconstructed nerve fibers, MorphoGNN captures the lower-dimensional representation of a single neuron and demonstrates cutting-edge performance in tasks such as neuron classification, retrieval, reconstruction quality classification, and neuron clustering.
IEEE TRANSACTIONS ON MEDICAL IMAGING
(2023)
Article
Biochemical Research Methods
Yufeng Liu, Ye Zhong, Xuan Zhao, Lijuan Liu, Liya Ding, Hanchuan Peng
Summary: In this study, a method called NeuMiner was proposed for tracing weak fibers by combining online sample mining strategy and modified gamma transformation. NeuMiner significantly improved the recall of weak signals, especially for axons.
Article
Computer Science, Interdisciplinary Applications
Andrea Behanova, Ali Abdollahzadeh, Ilya Belevich, Eija Jokitalo, Alejandra Sierra, Jussi Tohka
Summary: This study introduces a Matlab-based software called gACSON for analyzing myelinated axons in 3D electron microscopy (EM) imaging of brain tissue samples. The software features a graphical user interface, automatic segmentation of axonal space and myelin sheaths, and manual segmentation and interactive correction. The results suggest a decrease in axonal diameter after traumatic brain injury.
COMPUTER METHODS AND PROGRAMS IN BIOMEDICINE
(2022)
Article
Agriculture, Multidisciplinary
Eugene Kok, Xing Wang, Chao Chen
Summary: The shortage of agricultural labourers has led to the development of fruit harvesting robots globally. However, selective avoidance of obstacles such as tree branches in unstructured orchards remains a challenge. This study presents a framework that utilizes RGB-D camera data to reconstruct and recover obscured 3D branches, with promising results in branch segmentation and recovery accuracy.
COMPUTERS AND ELECTRONICS IN AGRICULTURE
(2023)
Article
Computer Science, Interdisciplinary Applications
Yi Jiang, Weixun Chen, Min Liu, Yaonan Wang, Erik Meijering
Summary: This paper proposes a neuronal structure segmentation method for 3D neuron microscopy images based on a combination of ray-shooting model and LSTM network, enhancing weak-signal neuronal structures and removing background noise. By transforming the 3D image segmentation task into multiple 1D ray/sequence segmentation tasks, the labeling of training samples is made much easier compared to existing CNN-based methods.
IEEE TRANSACTIONS ON MEDICAL IMAGING
(2021)
Article
Biochemical Research Methods
Christopher J. Handwerk, Katherine M. Bland, Collin J. Denzler, Anna R. Kalinowski, Cooper A. Brett, Brian D. Swinehart, Hilda Rodriguez, Hollyn N. Cook, Elizabeth C. Vinson, Madison E. Florenz, George S. Vidal
Summary: This study developed a method to simultaneously collect precise somatic positioning as well as 3D morphological data among transgenic fluorescent mouse hippocampal pyramidal neurons. This fluorescent method should be compatible with many other transgenic fluorescent reporter lines and immunohistochemical methods, facilitating the capture of topographic and morphological data from a wide variety of genetic experiments in mouse hippocampus.
JOURNAL OF NEUROSCIENCE METHODS
(2023)
Article
Computer Science, Software Engineering
Alejandro Beacco, Jaime Gallego, Mel Slater
Summary: This work presents an automatic method for 3D object reconstruction, which utilizes convolutional neural networks for semantic segmentation and warping of rendered depth maps to achieve high-quality results.
Article
Computer Science, Software Engineering
Attila Szabo, Georg Haaser, Harald Steinlechner, Andreas Walch, Stefan Maierhofer, Thomas Ortner, M. Eduard Groeller
Summary: This paper presents a framework for interactive, user-driven, feature-assisted geometry reconstruction from arbitrarily sized point clouds. It allows the user to extract planar pieces of geometry and utilize contextual suggestions to identify plane surfaces, directions, edges, and corners. The results are evaluated through systematic measurement of reconstruction accuracy and interviews with domain experts.
COMPUTERS & GRAPHICS-UK
(2023)
Article
Computer Science, Interdisciplinary Applications
Dongnan Liu, Donghao Zhang, Yang Song, Fan Zhang, Lauren O'Donnell, Heng Huang, Mei Chen, Weidong Cai
Summary: In this work, we propose a Panoptic Domain Adaptive Mask R-CNN (PDAM) method for unsupervised instance segmentation in microscopy images. By integrating semantic- and instance-level feature adaptation, our method effectively aligns cross-domain features at the panoptic level and solves domain bias issues. Experimental results demonstrate that the PDAM method outperforms state-of-the-art UDA methods by a large margin in various scenarios.
IEEE TRANSACTIONS ON MEDICAL IMAGING
(2021)
Article
Multidisciplinary Sciences
Priyanka Rana, Arcot Sowmya, Erik Meijering, Yang Song
Summary: The study introduces a 3D nuclear texture description method for cell nucleus classification and chromatin pattern variation measurement, which can aid in understanding cell differentiation, development, proliferation, and pathological conditions.
SCIENTIFIC REPORTS
(2021)
Article
Computer Science, Information Systems
Dongnan Liu, Chaoyi Zhang, Yang Song, Heng Huang, Chenyu Wang, Michael Barnett, Weidong Cai
Summary: Recent advances in unsupervised domain adaptation (UDA) techniques have successfully improved the generalization ability of data-driven deep learning architectures in cross-domain computer vision tasks. However, existing methods still face challenges in eliminating domain-specific factors from extracted features. To address this issue, we propose a Domain Disentanglement Faster-RCNN (DDF) method that disentangles both global and local features using GTD and ISD modules, respectively. Our DDF method outperforms state-of-the-art methods on four benchmark UDA object detection tasks, demonstrating its effectiveness and wide applicability.
IEEE TRANSACTIONS ON MULTIMEDIA
(2023)
Article
Chemistry, Analytical
Xuelin Zhang, Donghao Zhang, Alexander Leye, Adrian Scott, Luke Visser, Zongyuan Ge, Paul Bonnington
Summary: This paper focuses on improving the performance of scientific instrumentation that uses glass spray chambers for sample introduction, by detecting incidents using deep convolutional models. The indicators of poor quality sample introduction include the formation of liquid beads and flooding in the spray chamber. The proposed frameworks for detecting these incidents leverage modern deep learning architectures and expert knowledge, achieving high accuracy and real-time implementation.
Article
Computer Science, Information Systems
Yang Ma, Chaoyi Zhang, Mariano Cabezas, Yang Song, Zihao Tang, Dongnan Liu, Weidong Cai, Michael Barnett, Chenyu Wang
Summary: Multiple sclerosis (MS) is a chronic inflammatory and degenerative disease of the central nervous system, with magnetic resonance imaging (MRI) providing detailed structural information for the detection and categorization of MS lesions. Deep learning techniques have recently made significant breakthroughs in lesion segmentation, bypassing the need for manual annotation on 2D MRI slices.
IEEE JOURNAL OF BIOMEDICAL AND HEALTH INFORMATICS
(2022)
Article
Computer Science, Interdisciplinary Applications
Tiange Xiang, Yang Song, Chaoyi Zhang, Dongnan Liu, Mei Chen, Fan Zhang, Heng Huang, Lauren O'Donnell, Weidong Cai
Summary: We propose a novel weakly-supervised framework for classifying whole slide images (WSIs), which integrates information at both local and regional levels for better classification performance.
IEEE TRANSACTIONS ON MEDICAL IMAGING
(2022)
Article
Computer Science, Artificial Intelligence
Yicheng Wu, Zongyuan Ge, Donghao Zhang, Minfeng Xu, Lei Zhang, Yong Xia, Jianfei Cai
Summary: In this paper, a novel mutual consistency network (MC-Net+) is proposed for semi-supervised medical image segmentation. The MC-Net+ model effectively exploits un-labeled data and addresses the challenge of uncertain predictions in ambiguous regions. Experimental results demonstrate the superior performance of the proposed model compared to existing methods, establishing a new state-of-the-art for semi-supervised medical image segmentation.
MEDICAL IMAGE ANALYSIS
(2022)
Article
Computer Science, Interdisciplinary Applications
Lei Fan, Arcot Sowmya, Erik Meijering, Yang Song
Summary: In this paper, a novel framework for cancer prediction is proposed, which utilizes self-supervised learning methods to extract features from histopathological whole slide images and considers the overall survival of multiple patients. Experimental results demonstrate the excellent predictive accuracy of this framework.
IEEE TRANSACTIONS ON MEDICAL IMAGING
(2023)
Proceedings Paper
Computer Science, Artificial Intelligence
Yiwen Xu, Maurice Pagnucco, Yang Song
Summary: This paper proposes a Decoupled High-frequency semantic Guidance-based GAN (DHG-GAN) for diverse image outpainting, which aims to restore large missing regions surrounding a known region while generating multiple plausible results. Experimental results demonstrate that the proposed method outperforms existing approaches on CelebA-HQ, Place2, and Oxford Flower102 datasets.
COMPUTER VISION - ACCV 2022, PT VII
(2023)
Proceedings Paper
Computer Science, Artificial Intelligence
Zhihong Lin, Danli Shi, Donghao Zhang, Xianwen Shang, Mingguang He, Zongyuan Ge
Summary: Recent studies have confirmed the relationship between cardiovascular disease (CVD) risk and retinal fundus images. The combination of deep learning (DL) and portable fundus cameras can estimate CVD risk in various scenarios and improve healthcare availability. However, there are significant challenges, such as differences in cameras used in research databases and production environments.
MEDICAL IMAGE COMPUTING AND COMPUTER ASSISTED INTERVENTION, MICCAI 2022, PT II
(2022)
Proceedings Paper
Computer Science, Artificial Intelligence
Shek Wai Chu, Chaoyi Zhang, Yang Song, Weidong Cai
Summary: Human pose estimation is a challenging problem in computer vision, with recent advancements focusing on complex structure refinement and human joint graphical relations. Current research mainly aims at improving accuracy through iterative refinement subnetworks and exploitation of human joint graphical relations.
2021 IEEE INTERNATIONAL CONFERENCE ON IMAGE PROCESSING (ICIP)
(2021)
Proceedings Paper
Engineering, Biomedical
Cong Cong, Sidong Liu, Antonio Di Ieva, Maurice Pagnucco, Shlomo Berkovsky, Yang Song
Summary: Digital histopathology image analysis has become a hot research topic, with stain variation posing a significant challenge. This work introduces a novel approach using a texture enhanced pix2pix generative adversarial network to address stain normalization without the need for reference images, achieving excellent results in IDH mutation status classification.
2021 IEEE 18TH INTERNATIONAL SYMPOSIUM ON BIOMEDICAL IMAGING (ISBI)
(2021)
Proceedings Paper
Engineering, Biomedical
Priyanka Rana, Erik Meijering, Arcot Sowmya, Yang Song
Summary: This paper presents a multi-label classification pipeline and a novel feature descriptor for protein subcellular localization. By utilizing a Location-Sorted Random Projections feature descriptor and Multilabel Synthetic Minority Over-sampling Technique, the computational model achieves state-of-the-art performance on a highly unbalanced dataset with long-tail distribution and multi-label images. Additionally, the method shows excellent performance for minority classes.
2021 IEEE 18TH INTERNATIONAL SYMPOSIUM ON BIOMEDICAL IMAGING (ISBI)
(2021)
Article
Computer Science, Artificial Intelligence
Dongnan Liu, Donghao Zhang, Yang Song, Heng Huang, Weidong Cai
Summary: Instance segmentation in biomedical and biological image analysis is challenging due to complex backgrounds, variable object appearances, overlapping objects, and ambiguous boundaries. Proposed Panoptic Feature Fusion Net (PFFNet) unifies semantic and instance features to address the issue of information loss in proposal-free and proposal-based methods. PFFNet incorporates a residual attention feature fusion mechanism and mask quality sub-branch to improve semantic contextual information learning and align object confidence scores with mask quality prediction, leading to robust learning in both semantic and instance branches. Extensive experiments show PFFNet outperforms state-of-the-art methods on biomedical and biological datasets.
IEEE TRANSACTIONS ON IMAGE PROCESSING
(2021)