Zero-Shot Null-Space Diffusion Restoration with Adaptive Uncertainty-Guided Fusion for Ultrasound Speckle Reduction

Ultrasound B-mode imaging commonly suffers from speckle noise and artifacts, requiring a delicate balance between contrast, resolution, and preservation of anatomical structures. Although recently developed despeckling methods have achieved some progress, supervised learning approaches remain fundamentally limited by the ground truth paradox, which arises from the absence of noise-free, ground truth reference images in in vivo scenarios. Existing unsupervised diffusion-based methods typically enforce data consistency in the nonlinear log-compressed domain, which violates the physical additivity of acoustic signals and can disproportionately amplify background artifacts. To overcome these limitations, we propose an uncertainty-guided null-space diffusion solution (UGNS), a novel zero-shot framework that rigorously enforces consistency corrections in the physical linear domain. The proposed UGNS introduces several technical novelties: (a) extraction of a physically consistent structural prior through a dedicated structural prior to produce a robust signal envelope that preserves anatomical structure, (b) development of an adaptive range-null reconstruction mechanism that uses an adaptive weight mask to preserve tissue regions via range-space projection, and (c) introduction of uncertainty-guided fusion in an adaptive way to mitigate sampling variability. Extensive and comparative experiments were conducted using the PICMUS benchmark and in vivo datasets. The results demonstrate that our UGNS achieves better generalized contrast-to-noise ratios than other state-of-the-art methods. In addition, our UGNS effectively suppresses speckle noise while preserving fine spatial resolution.

Ultrasound Speckle ReductionUncertainty-Guided Null-Space DiffusionAdaptive Range-null-Space ReconstructionZero-Shot Diffusion Restoration

ADP-DiT: Text-Guided Diffusion Transformer for Brain Image Generation in Alzheimer’s Disease Progression

Alzheimer’s disease (AD) progresses heterogeneously across individuals, motivating subject-specific synthesis of follow-up magnetic resonance imaging (MRI) to support progression assessment. While Diffusion Transformers (DiT), an emerging transformer-based diffusion model, offer a scalable backbone for image synthesis, longitudinal AD MRI generation with clinically interpretable control over follow-up time and participant metadata remains underexplored. We present ADP-DiT, an interval-aware, clinically text-conditioned diffusion transformer for longitudinal AD MRI synthesis. ADP-DiT encodes follow-up interval together with multi-domain demographic, diagnostic (CN/MCI/AD), and neuropsychological information as a natural-language prompt, enabling time-specific control beyond coarse diagnostic stages. To inject this conditioning effectively, we use dual text encoders--OpenCLIP for vision–language alignment and T5 for richer clinical-language understanding. Their embeddings are fused into DiT through cross-attention for fine-grained guidance and adaptive layer normalization for global modulation. We further enhance anatomical fidelity by applying rotary positional embeddings to image tokens and performing diffusion in a pretrained SDXL-VAE latent space to enable efficient high-resolution reconstruction. On 3,321 longitudinal 3T T1-weighted scans from 712 participants (259,038 image slices), ADP-DiT achieves SSIM 0.8739 and PSNR 29.32 dB, improving over a DiT baseline by +0.1087 SSIM and +6.08 dB PSNR while capturing progression-related changes such as ventricular enlargement and shrinking hippocampus. These results suggest that integrating comprehensive, subject-specific clinical conditions with architectures can improve longitudinal AD MRI synthesis.

Alzheimer's DiseaseDisease ProgressionText-Guided Image GenerationDiffusion Transformer

Lightweight Temporal Segment Network for Video Scene Understanding: Validation in Driver Assault Detection

There has been an increasing number of driver assaults in transportation such as taxis and buses in the past years. It can be especially difficult to respond quickly to assaults on drivers by drunks late at night. To address this issue, our research team proposes a lightweight CNN-based Temporal Segment Network (TSN) model, which can detect driver assaults by passengers in real time. The TSN model efficiently processes videos by sampling a small number of image frames and is divided into two streams for learning: one for spatial information processing and the other for temporal information processing. Convolutional neural networks are employed in each stream. In this research, we apply a lightweight CNN architecture, MobileOne, significantly reducing the model size while demonstrating improved accuracy even with limited computing resources. The model is expected to contribute to a rapid response and prevention of hazardous situations for drivers by being integrated into vehicular driver monitoring systems.

Temporal Segment NetworkLightweight CNN ArchitectureAssault RecognitionAbnormal Behavior

Research on Developing a Responsive Web Interface for Small Businesses Using React

Although large corporations have established sales databases for efficient sales activities, small-scale entrepreneurs need a system that provides an interface that anyone can easily manage, offering diverse and complex information. This paper proposes the use of React technology, which can also operate on crawling servers that automatically collect store information. Furthermore, it suggests the development of a responsive web user interface that can be conveniently used in both mouse and keyboard-based desktop environments as well as touch-based smartphone environments, utilizing React technology for the effective delivery of various data. The proposed methods anticipate rapid retrieval of large volumes of data due to the adoption of asynchronous function processing in React technology, an improvement over traditional JavaScript methods.

Responsive Web InterfaceReactAsynchronous ProcessingInformation Architecture Design

Hankuk University of Foreign Studies

GPA: 4.37 / 4.5

Hankuk University of Foreign Studies

GPA: 3.50 / 4.5

SEOUL HIGH SCHOOL

Lightweight Ultrasound Blind Zone Restoration using Denoising Diffusion Restoration Models (DDRM)

Developed a lightweight Denoising Diffusion Restoration Model (DDRM) for real-time restoration of blind zones in ultrasound images (inference <15ms on edge devices). Achieved superior perceptual quality and improved SSIM/PSNR metrics by optimizing the diffusion process and model architecture for on-device efficiency.

Large-Scale Oceanographic Data Restoration for East Asia using Recurrent Feature Reasoning (RFR)

Successfully restored missing data in GOCI and UST21 satellite imagery across the entire East Asia region, expanding significantly beyond the initial Saemangeum and Nakdong River areas. Utilized a Recurrent Feature Reasoning (RFR) model for high-fidelity inpainting of Chlorophyll-a composite maps, enabling comprehensive environmental monitoring on a large geographical scale.

Application & Evaluation of DiffusionDet for Wave Overtopping Detection

Applied and fine-tuned the DiffusionDet model for real-world breakwater wave overtopping detection, demonstrating superior robustness and accuracy over the conventional YOLO model, especially under varying resolution and lighting conditions.

Wave Overtopping Detection & Tracking Experiment with YOLO v9 and ByteTrack

Upgraded object detection from YOLO v7 to YOLO v9 and replaced IOU Tracker with ByteTrack for wave overtopping. Demonstrated significant improvements in both detection and tracking accuracy (ID persistence) through rigorous quantitative comparison.