arXiv subjects

Yi Hong

Publications and source records attributed to Yi Hong.

At least 37 records · Page 2Linked to original sources

arXiv2024

Med3DInsight: Enhancing 3D Medical Image Understanding with 2D Multi-Modal Large Language Models

Understanding 3D medical image volumes is a critical task in the medical domain. However, existing 3D convolution and transformer-based methods have limited semantic understanding of an image volume and also need a large set of volumes for training. Recent advances in multi-modal large language models (MLLMs) provide a new and promising way to understand images with the help of text descriptions. However, most current MLLMs are designed for 2D natural images. To enhance the 3D medical image understanding with 2D MLLMs, we propose a novel pre-training framework called Med3DInsight, which marries existing 3D image encoders with 2D MLLMs and bridges them via a designed Plane-Slice-Aware Transformer (PSAT) module. Extensive experiments demonstrate our SOTA performance on two downstream segmentation and classification tasks, including three public datasets with CT and MRI modalities and comparison to more than ten baselines. Med3DInsight can be easily integrated into any current 3D medical image understanding network and improves its performance by a good margin.

cs.CV↗

arXiv2023

Longformer: Longitudinal Transformer for Alzheimer's Disease Classification with Structural MRIs

Structural magnetic resonance imaging (sMRI) is widely used for brain neurological disease diagnosis; while longitudinal MRIs are often collected to monitor and capture disease progression, as clinically used in diagnosing Alzheimer's disease (AD). However, most current methods neglect AD's progressive nature and only take a single sMRI for recognizing AD. In this paper, we consider the problem of leveraging the longitudinal MRIs of a subject for AD identification. To capture longitudinal changes in sMRIs, we propose a novel model Longformer, a spatiotemporal transformer network that performs attention mechanisms spatially on sMRIs at each time point and integrates brain region features over time to obtain longitudinal embeddings for classification. Our Longformer achieves state-of-the-art performance on two binary classification tasks of separating different stages of AD using the ADNI dataset. Our source code is available at https://github.com/Qybc/LongFormer.

cs.CV↗

arXiv2023

Volumetric Medical Image Segmentation via Scribble Annotations and Shape Priors

Recently, weakly-supervised image segmentation using weak annotations like scribbles has gained great attention in computer vision and medical image analysis, since such annotations are much easier to obtain compared to time-consuming and labor-intensive labeling at the pixel/voxel level. However, due to a lack of structure supervision on regions of interest (ROIs), existing scribble-based methods suffer from poor boundary localization. Furthermore, most current methods are designed for 2D image segmentation, which do not fully leverage the volumetric information if directly applied to each image slice. In this paper, we propose a scribble-based volumetric image segmentation, Scribble2D5, which tackles 3D anisotropic image segmentation and aims to its improve boundary prediction. To achieve this, we augment a 2.5D attention UNet with a proposed label propagation module to extend semantic information from scribbles and use a combination of static and active boundary prediction to learn ROI's boundary and regularize its shape. Also, we propose an optional add-on component, which incorporates the shape prior information from unpaired segmentation masks to further improve model accuracy. Extensive experiments on three public datasets and one private dataset demonstrate our Scribble2D5 achieves state-of-the-art performance on volumetric image segmentation using scribbles and shape prior if available.

cs.CV↗

arXiv2023

MedBLIP: Bootstrapping Language-Image Pre-training from 3D Medical Images and Texts

Vision-language pre-training (VLP) models have been demonstrated to be effective in many computer vision applications. In this paper, we consider developing a VLP model in the medical domain for making computer-aided diagnoses (CAD) based on image scans and text descriptions in electronic health records, as done in practice. To achieve our goal, we present a lightweight CAD system MedBLIP, a new paradigm for bootstrapping VLP from off-the-shelf frozen pre-trained image encoders and frozen large language models. We design a MedQFormer module to bridge the gap between 3D medical images and 2D pre-trained image encoders and language models as well. To evaluate the effectiveness of our MedBLIP, we collect more than 30,000 image volumes from five public Alzheimer's disease (AD) datasets, i.e., ADNI, NACC, OASIS, AIBL, and MIRIAD. On this largest AD dataset we know, our model achieves the SOTA performance on the zero-shot classification of healthy, mild cognitive impairment (MCI), and AD subjects, and shows its capability of making medical visual question answering (VQA). The code and pre-trained models is available online: https://github.com/Qybc/MedBLIP.

cs.CV↗

arXiv2023

MetaRegNet: Metamorphic Image Registration Using Flow-Driven Residual Networks

Deep learning based methods provide efficient solutions to medical image registration, including the challenging problem of diffeomorphic image registration. However, most methods register normal image pairs, facing difficulty handling those with missing correspondences, e.g., in the presence of pathology like tumors. We desire an efficient solution to jointly account for spatial deformations and appearance changes in the pathological regions where the correspondences are missing, i.e., finding a solution to metamorphic image registration. Some approaches are proposed to tackle this problem, but they cannot properly handle large pathological regions and deformations around pathologies. In this paper, we propose a deep metamorphic image registration network (MetaRegNet), which adopts time-varying flows to drive spatial diffeomorphic deformations and generate intensity variations. We evaluate MetaRegNet on two datasets, i.e., BraTS 2021 with brain tumors and 3D-IRCADb-01 with liver tumors, showing promising results in registering a healthy and tumor image pair. The source code is available online.

eess.IV↗

arXiv2022

Scribble2D5: Weakly-Supervised Volumetric Image Segmentation via Scribble Annotations

Recently, weakly-supervised image segmentation using weak annotations like scribbles has gained great attention, since such annotations are much easier to obtain compared to time-consuming and label-intensive labeling at the pixel/voxel level. However, because scribbles lack structure information of region of interest (ROI), existing scribble-based methods suffer from poor boundary localization. Furthermore, most current methods are designed for 2D image segmentation, which do not fully leverage the volumetric information if directly applied to image slices. In this paper, we propose a scribble-based volumetric image segmentation, Scribble2D5, which tackles 3D anisotropic image segmentation and improves boundary prediction. To achieve this, we augment a 2.5D attention UNet with a proposed label propagation module to extend semantic information from scribbles and a combination of static and active boundary prediction to learn ROI's boundary and regularize its shape. Extensive experiments on three public datasets demonstrate Scribble2D5 significantly outperforms current scribble-based methods and approaches the performance of fully-supervised ones. Our code is available online.

cs.CV↗

arXiv2022

Federated Learning Cost Disparity for IoT Devices

Federated learning (FL) promotes predictive model training at the Internet of things (IoT) devices by evading data collection cost in terms of energy, time, and privacy. We model the learning gain achieved by an IoT device against its participation cost as its utility. Due to the device-heterogeneity, the local model learning cost and its quality, which can be time-varying, differs from device to device. We show that this variation results in utility unfairness because the same global model is shared among the devices. By default, the master is unaware of the local model computation and transmission costs of the devices, thus it is unable to address the utility unfairness problem. Also, a device may exploit this lack of knowledge at the master to intentionally reduce its expenditure and thereby enhance its utility. We propose to control the quality of the global model shared with the devices, in each round, based on their contribution and expenditure. This is achieved by employing differential privacy to curtail global model divulgence based on the learning contribution. In addition, we devise adaptive computation and transmission policies for each device to control its expenditure in order to mitigate utility unfairness. Our results show that the proposed scheme reduces the standard deviation of the energy cost of devices by 99% in comparison to the benchmark scheme, while the standard deviation of the training loss of devices varies around 0.103.

cs.LG↗

arXiv2022

Terahertz Meets Untrusted UAV-Relaying: Minimum Secrecy Energy Efficiency Maximization via Trajectory and Communication Co-design

Unmanned aerial vehicles (UAVs) and Terahertz (THz) technology are envisioned to play paramount roles in next-generation wireless communications. In this paper, we present a novel secure UAV-assisted mobile relaying system operating at THz bands for data acquisition from multiple ground user equipments (UEs) towards a destination. We assume that the UAV-mounted relay may act, besides providing relaying services, as a potential eavesdropper called the untrusted UAV-relay (UUR). To safeguard end-to-end communications, we present a secure two-phase transmission strategy with cooperative jamming. Then, we devise an optimization framework in terms of a new measure $-$ secrecy energy efficiency (SEE), defined as the ratio of achievable average secrecy rate to average system power consumption, which enables us to obtain the best possible security level while taking UUR's inherent flight power limitation into account. For the sake of quality of service fairness amongst all the UEs, we aim to maximize the minimum SEE (MSEE) performance via the joint design of key system parameters, including UUR's trajectory and velocity, communication scheduling, and network power allocation. Since the formulated problem is a mixed-integer nonconvex optimization and computationally intractable, we decouple it into four subproblems and propose alternative algorithms to solve it efficiently via greedy/sequential block successive convex approximation and non-linear fractional programming techniques. Numerical results demonstrate significant MSEE performance improvement of our designs compared to other known benchmarks.

cs.IT↗

arXiv2022

Aerial Intelligent Reflecting Surface Enabled Terahertz Covert Communications in Beyond-5G Internet of Things

Unmanned aerial vehicles (UAVs) are envisioned to be extensively employed for assisting wireless communications in Internet of Things (IoT) applications. On the other hand, terahertz (THz) enabled intelligent reflecting surface (IRS) is expected to be one of the core enabling technologies for forthcoming beyond-5G wireless communications that promise a broad range of data-demand applications. In this paper, we propose a UAV-mounted IRS (UIRS) communication system over THz bands for confidential data dissemination from an access point (AP) towards multiple ground user equipments (UEs) in IoT networks. Specifically, the AP intends to send data to the scheduled UE, while unscheduled UEs may pose potential adversaries. To protect information messages and the privacy of the scheduled UE, we aim to devise an energy-efficient multi-UAV covert communication scheme, where the UIRS is for reliable data transmissions, and an extra UAV is utilized as a cooperative jammer generating artificial noise (AN) to degrade unscheduled UEs detection. We then formulate a novel minimum average energy efficiency (mAEE) optimization problem, targetting to improve the covert throughput and reduce UAVs' propulsion energy consumption subject to the covertness requirement, which is determined analytically. Since the optimization problem is non-convex, we tackle it via the block successive convex approximation (BSCA) approach to iteratively solve a sequence of approximated convex sub-problems, designing the binary user scheduling, AP's power allocation, maximum AN jamming power, IRS beamforming, and both UAVs' trajectory planning. Finally, we present a low-complex overall algorithm for system performance enhancement with complexity and convergence analysis. Numerical results are provided to verify our analysis and demonstrate significant outperformance of our design over other existing benchmark schemes.

cs.IT↗

arXiv2021

ASC-Net: Unsupervised Medical Anomaly Segmentation Using an Adversarial-based Selective Cutting Network

In this paper we consider the problem of unsupervised anomaly segmentation in medical images, which has attracted increasing attention in recent years due to the expensive pixel-level annotations from experts and the existence of a large amount of unannotated normal and abnormal image scans. We introduce a segmentation network that utilizes adversarial learning to partition an image into two cuts, with one of them falling into a reference distribution provided by the user. This Adversarial-based Selective Cutting network (ASC-Net) bridges the two domains of cluster-based deep segmentation and adversarial-based anomaly/novelty detection algorithms. Our ASC-Net learns from normal and abnormal medical scans to segment anomalies in medical scans without any masks for supervision. We evaluate this unsupervised anomly segmentation model on three public datasets, i.e., BraTS 2019 for brain tumor segmentation, LiTS for liver lesion segmentation, and MS-SEG 2015 for brain lesion segmentation, and also on a private dataset for brain tumor segmentation. Compared to existing methods, our model demonstrates tremendous performance gains in unsupervised anomaly segmentation tasks. Although there is still room to further improve performance compared to supervised learning algorithms, the promising experimental results and interesting observations shed light on building an unsupervised learning algorithm for medical anomaly identification using user-defined knowledge.

eess.IV↗

arXiv2021

Utility Fairness for the Differentially Private Federated Learning

Federated learning (FL) allows predictive model training on the sensed data in a wireless Internet of things (IoT) network evading data collection cost in terms of energy, time, and privacy. In this paper, for a FL setting, we model the learning gain achieved by an IoT device against its participation cost as its utility. The local model quality and the associated cost differs from device to device due to the device-heterogeneity which could be time-varying. We identify that this results in utility unfairness because the same global model is shared among the devices. In the vanilla FL setting, the master is unaware of devices' local model computation and transmission costs, thus it is unable to address the utility unfairness problem. In addition, a device may exploit this lack of knowledge at the master to intentionally reduce its expenditure and thereby boost its utility. We propose to control the quality of the global model shared with the devices, in each round, based on their contribution and expenditure. This is achieved by employing differential privacy to curtail global model divulgence based on the learning contribution. Furthermore, we devise adaptive computation and transmission policies for each device to control its expenditure in order to mitigate utility unfairness. Our results show that the proposed scheme reduces the standard deviation of the energy cost of devices by 99% in comparison to the benchmark scheme, while the standard deviation of the training loss of devices varies around 0.103.

cs.LG↗

arXiv2021

ASC-Net : Adversarial-based Selective Network for Unsupervised Anomaly Segmentation

We introduce a neural network framework, utilizing adversarial learning to partition an image into two cuts, with one cut falling into a reference distribution provided by the user. This concept tackles the task of unsupervised anomaly segmentation, which has attracted increasing attention in recent years due to their broad applications in tasks with unlabelled data. This Adversarial-based Selective Cutting network (ASC-Net) bridges the two domains of cluster-based deep learning methods and adversarial-based anomaly/novelty detection algorithms. We evaluate this unsupervised learning model on BraTS brain tumor segmentation, LiTS liver lesion segmentation, and MS-SEG2015 segmentation tasks. Compared to existing methods like the AnoGAN family, our model demonstrates tremendous performance gains in unsupervised anomaly segmentation tasks. Although there is still room to further improve performance compared to supervised learning algorithms, the promising experimental results shed light on building an unsupervised learning algorithm using user-defined knowledge.

cs.CV↗

arXiv2021

DDR-Net: Dividing and Downsampling Mixed Network for Diffeomorphic Image Registration

Deep diffeomorphic registration faces significant challenges for high-dimensional images, especially in terms of memory limits. Existing approaches either downsample original images, or approximate underlying transformations, or reduce model size. The information loss during the approximation or insufficient model capacity is a hindrance to the registration accuracy for high-dimensional images, e.g., 3D medical volumes. In this paper, we propose a Dividing and Downsampling mixed Registration network (DDR-Net), a general architecture that preserves most of the image information at multiple scales. DDR-Net leverages the global context via downsampling the input and utilizes the local details from divided chunks of the input images. This design reduces the network input size and its memory cost; meanwhile, by fusing global and local information, DDR-Net obtains both coarse-level and fine-level alignments in the final deformation fields. We evaluate DDR-Net on three public datasets, i.e., OASIS, IBSR18, and 3DIRCADB-01, and the experimental results demonstrate our approach outperforms existing approaches.

eess.IV↗

arXiv2021

Deterministic Pilot Design and Channel Estimation for Downlink Massive MIMO-OTFS Systems in Presence of the Fractional Doppler

Although the combination of the orthogonal time frequency space (OTFS) modulation and the massive multiple-input multiple-output (MIMO) technology can make communication systems perform better in high-mobility scenarios, there are still many challenges in downlink channel estimation owing to inaccurate modeling and high pilot overhead in practical systems. In this paper, we propose a channel state information (CSI) acquisition scheme for downlink massive MIMO-OTFS in presence of the fractional Doppler, including deterministic pilot design and channel estimation algorithm. First, we analyze the input-output relationship of the single-input single-output (SISO) OTFS based on the orthogonal frequency division multiplexing (OFDM) modem and extend it to massive MIMO-OTFS. Moreover, we formulate an accurate model for the practical system in which the fractional Doppler is considered and the influence of subpaths is revealed. A deterministic pilot design is then proposed based on the model and the structure of the pilot matrix to reduce pilot overhead and save memory consumption. Since channel geometry changes very slowly relative to the communication timescale, we put forward a modified sensing matrix based channel estimation (MSMCE) algorithm to acquire the downlink CSI. Simulation results demonstrate that the proposed downlink CSI acquisition scheme has significant advantages over traditional algorithms.

cs.IT↗

arXiv2021

MDA-Net: Multi-Dimensional Attention-Based Neural Network for 3D Image Segmentation

Segmenting an entire 3D image often has high computational complexity and requires large memory consumption; by contrast, performing volumetric segmentation in a slice-by-slice manner is efficient but does not fully leverage the 3D data. To address this challenge, we propose a multi-dimensional attention network (MDA-Net) to efficiently integrate slice-wise, spatial, and channel-wise attention into a U-Net based network, which results in high segmentation accuracy with a low computational cost. We evaluate our model on the MICCAI iSeg and IBSR datasets, and the experimental results demonstrate consistent improvements over existing methods.

cs.CV↗

arXiv2021

Orthogonal Time Sequency Multiplexing Modulation: Analysis and Low-Complexity Receiver Design

This paper proposes orthogonal time sequency multiplexing (OTSM), a novel single carrier modulation scheme that places information symbols in the delay-sequency domain followed by a cascade of time-division multiplexing (TDM) and Walsh-Hadamard sequence multiplexing. Thanks to the Walsh Hadamard transform (WHT), the modulation and demodulation do not require complex domain multiplications. For the proposed OTSM, we first derive the input-output relation in the delay-sequency domain and present a low complexity detection method taking advantage of zero-padding. We demonstrate via simulations that OTSM offers high performance gains over orthogonal frequency division multiplexing (OFDM) and similar performance to orthogonal time frequency space (OTFS), but at lower complexity owing to WHT. Then we propose a low complexity time-domain channel estimation method. Finally, we show how to include an outer error control code and a turbo decoder to improve error performance of the coded system.

eess.SP↗

arXiv2020

Low Complexity Precoding and Detection in Multi-user Massive MIMO OTFS Downlink

We consider the problem of degradation in performance of multi-carrier multi-user massive MIMO systems when channel induced Doppler spread is high. Recently, Orthogonal Time Frequency Space (OTFS) modulation has been shown to be robust to channel induced Doppler spread. In OTFS based systems, information symbols are embedded in the delay-Doppler (DD) domain where they are jointly modulated to generate the time-domain transmit signal. Due to the multi-path delay and Doppler shifts, the DD domain information symbols need to be jointly demodulated at the receiver. For multi-carrier based communication (e.g., Orthogonal Frequency Division Multiplexing (OFDM)), massive MIMO systems have been shown to achieve high spectral and energy efficiency with low complexity multi-user precoding in the downlink. Extending the same to OTFS based downlink multi-user massive MIMO systems is challenging due to the requirement for joint demodulation of all information symbols at the user terminal (UT). In this paper, we solve this problem by proposing a novel OTFS based multi-user precoder at the base station (BS) and a corresponding low complexity detector (LCD) at the user terminals (UTs), which allows for separate demodulation of each DD domain information symbol at the UT. The complexity of the proposed precoder increases only linearly with increasing number BS antennas Q and the number of UTs. We show, through analysis, that as Q increases (with total transmitted power decreased linearly with Q), the proposed low complexity detector achieves a sum SE close to that achieved with optimal joint demodulation at each UT. Numerical simulations confirm our analysis and also show that the SE and error rate performance of the proposed OTFS based massive MIMO precoder and LCD detector is significantly more robust to channel induced Doppler spread when compared to OFDM based multi-user massive MIMO systems.

cs.IT↗

arXiv2020

Joint Trajectory and Power Allocation Design for Secure Artificial Noise aided UAV Communications

This paper investigates an average secrecy rate (ASR) maximization problem for an unmanned aerial vehicle (UAV) enabled wireless communication system, wherein a UAV is employed to deliver confidential information to a ground destination in the presence of a terrestrial passive eavesdropper. By employing an artificial noise (AN) injection based secure two-phase transmission protocol, we aim at jointly optimizing the UAV's trajectory, network transmission power, and AN power allocation over a given time horizon to enhance the ASR performance. Specifically, we divide the original non-convex problem into four subproblems, and propose a successive convex approximation based efficient iterative algorithm to solve it suboptimally with guaranteed convergence. Simulation results demonstrate significant security advantages of our designed scheme over other known benchmarks, particularly for stringent flight durations.

cs.IT↗