Heung‐Il Suk

Causal Manifold Transport for Identifiable Causal Generation in Diffusion Models

Junghyo Sohn, Wootaek Jeong, Sujeong Song, Jee Seok Yoon et autres

Identifying meaningful latent representations within diffusion models remains a challenging problem for causal approaches. We propose Causal Manifold Transport Diffusion Model (CMT-Diff), a framework that operationalizes causal actions as geometric transformations. By adopting the perspective of backtracking counterfactuals, we formulate the generative …

jp, kr (code pays fourni par la source)

DyMix: Dynamic frequency Mixup scheduler based unsupervised domain adaptation for enhancing Alzheimer’s disease prediction

Kwanseok Oh, Yooseung Shin, Heung‐Il Suk

Recent advances in deep learning (DL) have substantially improved the accuracy of Alzheimer’s disease (AD) diagnosis from brain images, enabling earlier and more reliable clinical interventions. Nevertheless, most DL-based models often suffer from significant performance degradation when applied to unseen domains owing …

kr (code pays fourni par la source)

STRIVE: Multi-Agent Structured Temporal Reasoning with Integrated Verification for Longitudinal Radiology Report Generation

Junyeong Maeng, Eunsong Kang, Heung‐Il Suk

Longitudinal radiology report generation (LRRG) requires identifying both current findings and their changes relative to a prior study. Existing methods jointly model diagnosis, attribute estimation, temporal comparison, and language generation within implicit representations, which can cause task interference, obscure the evidence underlying …

SAM-TXD: SAM with Text-Guided Diffusion for Refined Medical Image Segmentation

Seunghee Yoo, Junghyo Sohn, Kwanseok Oh, Heung‐Il Suk

While the Segment Anything Model (SAM) has spurred extensive research in medical image segmentation, its direct application to medical images is often suboptimal, struggling with ambiguous boundaries and limited semantic context. We propose SAM-TXD, a diffusion-based framework that refines SAM outputs by …

kr (code pays fourni par la source)

PAT-DIFF: Pathology-Aware Residual Diffusion Framework for MRI-to-PET Translation in Alzheimer's Disease

Ha Eun Kim, Kwanseok Oh, Heung‐Il Suk

Positron Emission Tomography (PET) and Magnetic Resonance Imaging (MRI) provide complementary information for Alzheimer's disease (AD) diagnosis, but acquiring both is often impractical owing to the high cost and radiation exposure associated with PET. To bridge these complementary domains, we propose PaT-Diff, …

kr (code pays fourni par la source)

HyFI: Hyperbolic Feature Interpolation for Brain-Vision Alignment

Sangmin Jo, Wootaek Jeong, Da-Woon Heo, Yoohwan Hwang et autres

Recent progress in artificial intelligence has encouraged numerous attempts to understand and decode human visual system from brain signals. These prior works typically align neural activity independently with semantic and perceptual features extracted from images using pre-trained vision models. However, they fail …

SIGMAformer: a spatiotemporal Gaussian mixture correlation transformer for global weather forecasting

D. H. Kim, Heung‐Il Suk

Abstract SpatIotemporal Gaussian Mixture correlAtion transformer (SIGMAformer) is a spatiotemporal forecasting architecture that integrates a Gaussian mixture pattern extractor (GMPE) with a dynamic spatiotemporal correlation (DSTC) mechanism. The DSTC module leverages GMPE to automatically compute spatiotemporal pattern-specific weights from the data. These …

kr (code pays fourni par la source)

HyFI: Hyperbolic Feature Interpolation for Brain-Vision Alignment

Sang‐min Jo, Wootaek Jeong, Da-Woon Heo, Yoohwan Hwang et autres

Recent progress in artificial intelligence has encouraged numerous attempts to understand and decode human visual system from brain signals. These prior works typically align neural activity independently with semantic and perceptual features extracted from images using pre-trained vision models. However, they fail …

Permutation-Equivariant 2D State Space Models: Theory and Canonical Architecture for Multivariate Time Series

Seungwoo Jeong, Heung‐Il Suk

Multivariate time series (MTS) modeling often implicitly imposes an artificial ordering over variables, violating the inherent exchangeability found in many real-world systems where no canonical variable axis exists. We formalize this limitation as a violation of the permutation symmetry principle and require …

Permutation-Equivariant 2D State Space Models: Theory and Canonical Architecture for Multivariate Time Series

Seungwoo Jeong, Heung‐Il Suk

Multivariate time series (MTS) modeling often implicitly imposes an artificial ordering over variables, violating the inherent exchangeability found in many real-world systems where no canonical variable axis exists. We formalize this limitation as a violation of the permutation symmetry principle and require …

Transferring ultrahigh-field representations for intensity-guided brain segmentation of low-field magnetic resonance imaging

Kwanseok Oh, Jieun Lee, Da-Woon Heo, Dinggang Shen et autres

Ultrahigh-field (UHF) magnetic resonance imaging (MRI), 7T MRI, provides superior anatomical details of internal brain structures thanks to its enhanced signal-to-noise ratio and susceptibility-induced contrast. However, the widespread use of 7T MRI is limited by its high cost and lower accessibility compared …

kr, cn (code pays fourni par la source)

Why Go Heavy? Rethinking Visual Information Reduction in EEG-to-Image Alignment

Sangmin Jo, Adrián Ayuso-Muñoz, Heung‐Il Suk

kr (code pays fourni par la source)