BiX

2025

MARSeg: Enhancing Medical Image Segmentation with MAR and Adaptive Feature Fusion

Jeonghyun Hwang, Seungyeon Rhee, Minjeong Kim, Thanaporn Viriyasaranon, Jang-Hwan Choi

MICCAI 2025 (Oral)

MARSeg combines metal artifact reduction and adaptive feature fusion to improve medical image segmentation in artifact-degraded scans, supporting more robust downstream analysis.

Unsupervised motion artifacts reduction for cone-beam CT via enhanced landmark detection

Thanaporn Viriyasaranon, Serie Ma, Mareike Thies, Andreas Maier, Jang-Hwan Choi

Expert Systems with Applications, 278:127258 (2025)

A landmark-driven motion estimation framework reduces motion artifacts in cone-beam CT without motion-free reference scans or external tracking hardware.

VIEW PUBLICATION

Unsupervised Training of a Dynamic Context-Aware Deep Denoising Framework for Low-Dose Fluoroscopic Imaging

Sun-Young Jeon, Sen Wang, Adam S. Wang, Garry E. Gold, Jang-Hwan Choi

IEEE Transactions on Instrumentation and Measurement, 74, 1-15 (2025)

An unsupervised temporal framework suppresses correlated and uncorrelated noise in low-dose fluoroscopy while adapting to object motion and preserving clinically important edges.

VIEW PUBLICATION

Low-dose computed tomography perceptual image quality assessment

Wonkyeong Lee, Fabian Wagner, Adrian Galdran, Yongyi Shi, Wenjun Xia, Ge Wang, Xuanqin Mou, Md Atik Ahamed, Abdullah Al Zubaer Imran, Ji Eun Oh, Kyungsang Kim, Jong Tak Baek, Dongheon Lee, Boohwi Hong, Philip Tempelman, Donghang Lyu, Adrian Kuiper, Lars van Blokland, Maria Baldeon Calisto, Scott Hsieh, Minah Han, Jongduk Baek, Andreas Maier, Adam Wang, Garry Evan Gold, Jang-Hwan Choi

Medical Image Analysis, 99:103343 (2025)

The work establishes a radiologist-scored low-dose CT quality dataset and benchmark, showing how reference-free AI metrics can better reflect clinically perceived image quality.

VIEW PUBLICATION

2024

A systematic review of deep learning-based denoising for low-dose computed tomography from a perceptual quality perspective

Wonjin Kim, Sun-Young Jeon, Gyuri Byun, Hongki Yoo, Jang-Hwan Choi

Biomedical Engineering Letters, 14(6), 1153-1173 (2024)

This review evaluates low-dose CT denoising through perceptual and diagnostic quality, highlighting the limitations of conventional numerical metrics and overly smoothed outputs.

VIEW PUBLICATION

An Unsupervised Two-step Training Framework for Low-dose Computed Tomography Denoising

Wonjin Kim, Jaayeon Lee, Jang-Hwan Choi

Medical Physics, 51(2), 1127-1144 (2024)

A two-stage framework combines volumetric self-learning with a memory-efficient denoising GAN to improve both objective fidelity and perceived quality without paired clinical scans.

VIEW PUBLICATION

2023

Annotation-Efficient Deep Learning Model for Pancreatic Cancer Diagnosis and Classification Using CT Images: A Retrospective Diagnostic Study

Thanaporn Viriyasaranon, Jung Won Chun, Young Hwan Koh, Jae Hee Cho, Min Kyu Jung, Seong-Hun Kim, Hyo Jung Kim, Woo Jin Lee, Jang-Hwan Choi, Sang Myung Woo

Cancers, 15(13), 3392 (2023)

A self-supervised pseudo-lesion segmentation strategy improves pancreatic cancer classification from CT images while reducing dependence on extensive expert annotations and retaining performance on external patient data.

VIEW PUBLICATION

A Multi-kernel and Multi-scale Learning Based Deep Ensemble Model for Predicting Recurrence of Non-small Cell Lung Cancer

Gihyeon Kim, Young Mi Park, Hyun Jung Yoon, Jang-Hwan Choi

PeerJ Computer Science, 9:e1311 (2023)

The study combines multiple CT slices, spatial scales, and convolution kernels to estimate recurrence risk in patients with non-small cell lung cancer and support treatment planning.

VIEW PUBLICATION

MM-Net: Multiframe and Multimask-Based Unsupervised Deep Denoising for Low-Dose Computed Tomography

Sun-Young Jeon, Wonjin Kim, Jang-Hwan Choi

IEEE Transactions on Radiation and Plasma Medical Sciences, 7(3), 296-306 (2023)

MM-Net learns volumetric low-dose CT denoising from neighboring slices and masked patches without paired normal-dose ground truth.

VIEW PUBLICATION

2022

No-reference perceptual CT image quality assessment based on a self-supervised learning framework

Wonkyeong Lee, Eunbyeol Cho, Wonjin Kim, Hyebin Choi, Kyongmin Sarah Beck, Hyun Jung Yoon, Jongduk Baek, Jang-Hwan Choi

Machine Learning: Science and Technology, 3(4):045033 (2022)

A self-supervised model estimates radiologist-aligned CT image quality without requiring a pristine reference image, supporting automated quality control in clinical imaging.

VIEW PUBLICATION

Wavelet subband-specific learning for low-dose computed tomography denoising

Wonjin Kim, Jaayeon Lee, Mihyun Kang, Jin Sung Kim, Jang-Hwan Choi

PLOS ONE, 17(9):e0274308 (2022)

A wavelet-domain network separately optimizes low- and high-frequency information to improve noise reduction while retaining realistic texture and structural detail in low-dose CT.

VIEW PUBLICATION