Predicting time-to-conversion for dementia of Alzheimer's type using multi-modal deep survival analysis

Article Computer Science, Interdisciplinary Applications

Deep Bayesian Hashing With Center Prior for Multi-Modal Neuroimage Retrieval

Erkun Yang, Mingxia Liu, Dongren Yao, Bing Cao, Chunfeng Lian, Pew-Thian Yap, Dinggang Shen

Summary: The study introduces a deep Bayesian hash learning framework called CenterHash, which maps multi-modal neuroimage data into a shared Hamming space and learns discriminative hash codes from imbalanced neuroimages.

IEEE TRANSACTIONS ON MEDICAL IMAGING (2021)

Add to Collection

Article Chemistry, Multidisciplinary

Integrative Serum Metabolic Fingerprints Based Multi-Modal Platforms for Lung Adenocarcinoma Early Detection and Pulmonary Nodule Classification

Lin Wang, Mengji Zhang, Xufeng Pan, Mingna Zhao, Lin Huang, Xiaomeng Hu, Xueqing Wang, Lihua Qiao, Qiaomei Guo, Wanxing Xu, Wenli Qian, Tingjia Xue, Xiaodan Ye, Ming Li, Haixiang Su, Yinglan Kuang, Xing Lu, Xin Ye, Kun Qian, Jiatao Lou

Summary: A multiplexed assay is developed on a nanoparticle-based laser desorption/ionization mass spectrometry platform for the detection of serum metabolic fingerprints (SMFs) in lung adenocarcinoma (LUAD). A dual modal model, MP-NN, integrating SMFs with protein tumor marker CEA via deep learning, shows superior performance compared to a single modal model. The tri modal model, MPI-RF, integrating SMFs, tumor marker CEA, and image features, demonstrates significantly higher performance in pulmonary nodule classification.

ADVANCED SCIENCE (2022)

Add to Collection

Article Computer Science, Theory & Methods

A Comprehensive Report on Machine Learning-based Early Detection of Alzheimer's Disease using Multi-modal Neuroimaging Data

Shallu Sharma, Pravat Kumar Mandal

Summary: This paper outlines a machine learning approach for early diagnosis of Alzheimer's Disease using multi-modal neuroimaging data. By extracting and selecting features, as well as scaling and fusing data, an ML-based diagnosis system can be designed. Additionally, thematic analysis is provided to compare the ML workflow for different diagnostic solutions.

ACM COMPUTING SURVEYS (2023)

Add to Collection

Article Computer Science, Artificial Intelligence

MDMN: Multi-task and Domain Adaptation based Multi-modal Network for early rumor detection

Honghao Zhou, Tinghuai Ma, Huan Rong, Yurong Qian, Yuan Tian, Najla Al-Nabhan

Summary: With the growing popularity of social media, people are increasingly expressing their opinions through multimedia content. This paper proposes a Multi-task and Domain Adaptation based Multi-modal Network (MDMN) to improve the accuracy of rumor detection in multi-modal data. The network includes components such as Textual Feature Extractor, Visual Feature Extractor, and Fusion & Classification Network, and uses task-specific methods to enhance the representation of textual data. The experiment shows that MDMN outperforms baseline methods, achieving a recall rate of over 92%.

EXPERT SYSTEMS WITH APPLICATIONS (2022)

Add to Collection

Article Clinical Neurology

Higher performance for women than men in MRI-based Alzheimer′s disease detection

Malte Klingenberg, Didem Stark, Fabian Eitel, Celine Budding, Mohamad Habes, Kerstin Ritter, Alzheimers Dis Neuroimaging Initiat

Summary: This study trained a convolutional neural network using a balanced dataset to detect Alzheimer's disease. The results showed that the machine learning classifier had different performance for men and women, indicating the presence of sex bias. The findings emphasize the importance of examining and reporting classifier performance across population subgroups to ensure algorithmic fairness.

ALZHEIMERS RESEARCH & THERAPY (2023)

Add to Collection

Article Nanoscience & Nanotechnology

Dumbbell Aptamer Sensor Based on Dual Biomarkers for Early Detection of Alzheimer?s Disease

Jie Zhou, Yiwen Sun, Jin Zhang, Fusui Luo, Huili Ma, Min Guan, Junfen Feng, Xiaomeng Dong

Summary: In this study, a detection system based on the entropy-driven strand displacement reaction (ESDR) principle was developed for the sensitive detection of miR-193b and AflO42, biomarkers for early stage AD. The system consisted of a dumbbell detection probe (H), an indicator probe (R), and graphene oxide (GO). GO adsorbed free R and quenched fluorescence, enhancing the sensitivity of the system. The detection limits for miR-193b and AflO42 were 77 pM and 53 pM, respectively.

ACS APPLIED MATERIALS & INTERFACES (2023)

Add to Collection

Article Nanoscience & Nanotechnology

Dumbbell Aptamer Sensor Based on Dual Biomarkers for Early Detection of Alzheimer?s Disease

Jie Zhou, Yiwen Sun, Jin Zhang, Fusui Luo, Huili Ma, Min Guan, Junfen Feng, Xiaomeng Dong

Summary: Finding a timely, sensitive, and noninvasive detection method for early diagnosis of Alzheimer's disease (AD) has become urgent. MicroRNA-193b (miR-193b) and Afl42 oligomers (AflO42) in neurogenic exosomes were identified as biomarkers reflecting pathological changes in the early stage of AD. A detection system based on the entropy-driven strand displacement reaction (ESDR) principle was developed, showing high specificity, sensitivity, and ease of operation, providing broad prospects for early diagnosis of AD.

ACS APPLIED MATERIALS & INTERFACES (2023)

Add to Collection

Review Biochemistry & Molecular Biology

Deep Learning for Genomics: From Early Neural Nets to Modern Large Language Models

Tianwei Yue, Yuanxin Wang, Longxiang Zhang, Chunming Gu, Haoru Xue, Wenping Wang, Qi Lyu, Yujie Dun

Summary: This paper briefly discusses the strengths of different deep learning models from a genomic perspective and comments on the practical considerations of developing deep learning architectures for genomics. It also provides a concise review of deep learning applications in various aspects of genomic research and points out current challenges and potential research directions for future genomics applications. The collaborative use of ever-growing diverse data and the fast iteration of deep learning models are believed to contribute to the future of genomics.

INTERNATIONAL JOURNAL OF MOLECULAR SCIENCES (2023)

Add to Collection

Article Computer Science, Information Systems

Correspondence Learning for Deep Multi-Modal Recognition and Fraud Detection

Jongchan Park, Min-Hyun Kim, Dong-Geol Choi

Summary: This study proposes a correspondence learning technique to explicitly learn the relationship among multiple modalities, achieving better representations in multi-modal recognition tasks. The method is validated on various multi-modal benchmarks and also demonstrates a fraud detection method using the learned correspondence among modalities.

M 2RNet: Multi-modal and multi-scale refined network for RGB-D salient object detection

Xian Fang, Mingfeng Jiang, Jinchao Zhu, Xiuli Shao, Hongpeng Wang

Summary: This paper presents a novel multi-modal and multi-scale refined network to address the challenges of multi-modal feature fusion and multi-scale feature aggregation in RGB-D images. The proposed network achieves superior performance compared to state-of-the-art approaches, as demonstrated by extensive quantitative and qualitative experiments.

PATTERN RECOGNITION (2023)

Add to Collection

Article Computer Science, Artificial Intelligence

TRIMOON: Two-Round Inconsistency-based Multi-modal fusion Network for fake news detection

Shufeng Xiong, Guipei Zhang, Vishwash Batra, Lei Xi, Lei Shi, Liangliang Liu

Summary: Compared to ordinary news, fake news spreads faster with lower production cost, causing significant social harm. Detecting fake news efficiently and accurately has become a research focus due to these reasons. We propose a Two-Round Inconsistency-based Multi-modal fusion Network (TRIMOON) for fake news detection, consisting of feature extraction, fusion, and classification modules. By performing two-fold inconsistency detection, we effectively filter noise generated during the fusion process. Experimental results demonstrate the superiority of our TRIMOON model over state-of-the-art approaches on Chinese and English datasets.

INFORMATION FUSION (2023)

Add to Collection

Article Computer Science, Information Systems

Deep learning based object detection from multi-modal sensors: an overview

Ye Liu, Shiyang Meng, Hongzhang Wang, Jun Liu

Summary: Object detection is an important problem with a wide range of applications, and recent progress has been made in deep learning based object detection using RGB cameras. However, the limitations of RGB cameras are becoming more apparent. Additional sensors on unmanned vehicles or mobile robot platforms can expand the sensing range of RGB cameras from different dimensions. This paper summarizes deep learning based object detection methods under the condition of multi-modal sensors and categorizes them from the perspective of data fusion.

MULTIMEDIA TOOLS AND APPLICATIONS (2023)

Add to Collection

Article Computer Science, Artificial Intelligence

MMHFNet: Multi-modal and multi-layer hybrid fusion network for voice pathology detection

Hussein M. A. Mohammed, Asli Nur Omeroglu, Emin Argun Oral

Summary: This paper presents a novel deep Multi-Modal and Multi-Layer Hybrid Fusion Network (MMHFNet) for improving the performance of non-invasive voice pathology detection systems. MMHFNet combines complementary information from different modalities (speech and EGG signals). It vertically combines low-level and high-level features to take advantage of the spatio-spectral information for multi-layer fusion. The extracted features are then fed into an LSTM classification network for voice pathology diagnosis.

EXPERT SYSTEMS WITH APPLICATIONS (2023)

Add to Collection

Article Biochemical Research Methods

Hyper-graph based sparse canonical correlation analysis for the diagnosis of Alzheimer?s disease from multi-dimensional genomic data

Wei Shao, Shunian Xiang, Zuoyi Zhang, Kun Huang, Jie Zhang

Summary: The diagnosis of Alzheimer's disease (AD), especially in the early stage, remains a challenge in research. Multiple biomarkers have been associated with AD diagnosis, but existing research often only uses single modality data. This study proposes a method that integrates multi-modal genomic data to extract biomarkers associated with AD and MCI.

METHODS (2021)

Add to Collection

Article Environmental Sciences

SymmetricNet: end-to-end mesoscale eddy detection with multi-modal data fusion

Yuxiao Zhao, Zhenlin Fan, Haitao Li, Rui Zhang, Wei Xiang, Shengke Wang, Guoqiang Zhong

Summary: In this paper, an end-to-end mesoscale eddy detection method based on multi-modal data fusion is proposed. Existing methods using single-modal data such as sea surface height (SSH) for detection result in inaccurate results. The proposed method not only uses SSH, but also includes data from other modalities such as sea surface temperature (SST) and velocity of flow. Moreover, a novel network named SymmetricNet is designed to achieve multi-modal data fusion in mesoscale eddy detection.

FRONTIERS IN MARINE SCIENCE (2023)

Add to Collection

Article Mathematical & Computational Biology

Predicting the onset of breast cancer using mammogram imaging data with irregular boundary

Shu Jiang, Jiguo Cao, Graham A. Colditz, Bernard Rosner

Summary: With mammography as the primary screening strategy for breast cancer, it is important to utilize mammogram imaging data to identify women at different risk levels. This study proposes a supervised functional principal component analysis method for extracting features from mammogram images, which improves prediction accuracy. The method effectively addresses the irregular boundary issue and shows better performance compared to unsupervised methods in simulation studies.

BIOSTATISTICS (2023)

Add to Collection

Article Statistics & Probability

A Dynamic Interaction Semiparametric Function-on-Scalar Model

Hua Liu, Jinhong You, Jiguo Cao

Summary: Motivated by recent work on massive functional data, this study proposes a new dynamic interaction semiparametric function-on-scalar (DISeF) model, which is useful to explore the dynamic interaction among a set of covariates and their effects on the functional response. The study develops an efficient estimation procedure to iteratively estimate the bivariate varying-coefficient functions, the index parameters, and the covariance functions of random effects. It also establishes the asymptotic properties of the estimators and develops a test statistic to check the variability of the dynamic interaction.

JOURNAL OF THE AMERICAN STATISTICAL ASSOCIATION (2023)

Add to Collection

Article Clinical Neurology

AAV5-miHTT-mediated huntingtin lowering improves brain health in a Huntington's disease mouse model

Sarah B. Thomson, Anouk Stam, Cynthia Brouwers, Valentina Fodale, Alberto Bresciani, Michael Vermeulen, Sara Mostafavi, Terri L. Petkau, Austin Hill, Andrew Yung, Bretta Russell-Schulz, Piotr Kozlowski, Alex MacKay, Da Ma, Mirza Faisal Beg, Melvin M. Evers, Astrid Valles, Blair R. Leavitt

Summary: This study investigates the effects of AAV5-miHTT treatment in a mouse model of Huntington's disease, and shows that it can improve brain health and reverse transcriptional dysregulation in the model.

BRAIN (2023)

Add to Collection

Article Environmental Sciences

Nonlinear prediction of functional time series

Haixu Wang, Jiguo Cao

Summary: We propose a nonlinear prediction (NOP) method for functional time series. This method addresses the limitations of conventional approaches, which rely on the stationary or linear assumption of the functional time series and struggle with multivariate functional time series. The NOP method employs a nonlinear mapping for functional data, avoids calculating covariance functions, and enables online estimation and prediction, showing superior prediction performances in simulations and real applications.

ENVIRONMETRICS (2023)

Add to Collection

Article Environmental Sciences

Estimating functional single index models with compact support

Yunlong Nie, Liangliang Wang, Jiguo Cao

Summary: Functional single index models are commonly used to describe the nonlinear relationship between a scalar response and a functional predictor. We propose a new compact functional single index model that can identify the region in which the functional predictor is related to the response. The effectiveness of our method is demonstrated through an application example and a simulation study.

ENVIRONMETRICS (2023)

Add to Collection

Article Computer Science, Theory & Methods

Automatic search intervals for the smoothing parameter in penalized splines

Zheyuan Li, Jiguo Cao

Summary: The selection of smoothing parameter is crucial for penalized splines estimation. We have developed algorithms to automatically find the optimal smoothing parameter range, which has four advantages.

STATISTICS AND COMPUTING (2023)

Add to Collection

Article Environmental Sciences

Organotin Antifouling Compounds and Sex-Steroid Nuclear Receptor Perturbation: Some Structural Insights

Mohd A. Beg, Md A. Beg, Ummer R. Zargar, Ishfaq A. Sheikh, Osama S. Bajouh, Adel M. Abuzenadah, Mohd Rehan

Summary: Organotin compounds (OTCs) are widely used as polyvinyl chloride stabilizers and marine antifouling biocides. They have metabolic and endocrine disrupting effects in organisms and may interfere with natural steroid/receptor binding and perturb steroid signaling.

TOXICS (2023)

Add to Collection

Article Biology

Segmentation-guided domain adaptation and data harmonization of multi-device retinal optical coherence tomography using cycle-consistent generative adversarial networks

Shuo Chen, Da Ma, Sieun Lee, Timothy T. L. Yu, Gavin Xu, Donghuan Lu, Karteek Popuri, Myeong Jin Ju, Marinko Sarunic, Mirza Faisal Beg

Summary: This study introduces a segmentation-driven domain adaptation method for retinal imaging processing, which utilizes CycleGAN to improve domain adaptation. A two-stage pipeline is proposed to train a segmentation model in the source domain and adapt the images from the target domain. Experimental results demonstrate that the proposed method achieves improved segmentation performance and image quality.

COMPUTERS IN BIOLOGY AND MEDICINE (2023)

Add to Collection

Article Neurosciences

Longitudinal Spatial Relationships Between Atrophy and Hypometabolism Across the Alzheimer's Disease Continuum

Jane Stocks, Ashley Heywood, Karteek Popuri, Mirza Faisal Beg, Howie Rosen, Lei Wang

Summary: By analyzing the multimodal neuroimaging relationships between MRI and 18FDG-PET, it was found that in the suspected non-Alzheimer's disease pathology (SNAP) group, there is a spatially overlapping relationship between brain atrophy and hypometabolism at the M-12 timepoint. In the Amyloid Only group, there is a spatially discordant relationship between distributed atrophy and hypometabolism at all time points. In Probable AD subjects, there is a local correlation between bilateral temporal lobes at baseline and M-12 when both modalities are assessed. Across groups, hypometabolism at baseline is correlated with non-local atrophy at M-12. These results support the view that local concordance of atrophy and hypometabolism is the result of a tau-mediated process driving neurodegeneration.

JOURNAL OF ALZHEIMERS DISEASE (2023)

Add to Collection

Article Computer Science, Artificial Intelligence

A moment-matching metric for latent variable generative models

Cedric Beaulac

Summary: Assessing the quality of a fitted model is challenging in unsupervised learning. We propose a new metric based on moments to compare and regularize models. We demonstrate the applications of this metric and discuss future research directions.

MACHINE LEARNING (2023)

Add to Collection

Article Health Care Sciences & Services

Identifying regions of interest in mammogram images

Shu Jiang, Jiguo Cao, Graham A. A. Colditz

Summary: Screening mammography is crucial for early detection and prevention of breast cancer, but the irregular boundary of breast area in mammograms poses challenges in identifying risk-associated regions. We propose a proportional hazards model with imaging predictors characterized by bivariate splines over triangulation, enforced with group lasso penalty function, to address these challenges and achieve higher discriminatory performance.

STATISTICAL METHODS IN MEDICAL RESEARCH (2023)

Add to Collection

Article Statistics & Probability

Jointly modelling multiple transplant outcomes by a competing risk model via functional principal component analysis

Jianghu (James) Dong, Haolun Shi, Liangliang Wang, Ying Zhang, Jiguo Cao

Summary: Longitudinal biomarkers are commonly used in clinical studies to monitor disease progression. This study develops a joint model that utilizes functional principal component analysis to extract features from longitudinal trajectories and employs a competing risk model to handle multiple time-to-event outcomes.

JOURNAL OF APPLIED STATISTICS (2023)

Add to Collection