_tleeiln _ienr26o rtrnb:..::iI ::::::::_anf nfn:_6M::..::e: ::::::0IDLAg_tl i_:.::oI_G_n=-=-=- =-=-=-::amNr::::bi gg_:::::::-=-=-=-=-= -=-=-=-=-=::::...:sp an:::::::-=-=-=-=-=-=- =-=-=-=-=-=-=:::::::_i i_::::m-=-=-=-=-=-=-=-= -=-=-=-=-=-=-=-=:rEE__u g:nO2o:-=-=-=-=-=-=-=-=- =-=-=-=-=-=-=-=-=:o2Oi_b f.:0i:=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-:::::a _o:..::=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-::..:nf t_::::=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-:::::i e2OI_G=-=-=-=-=-=-=-=((-=((-=-=))-=))-=-=-=-=-=-=-=-:i_E_l _tdLAG=-=-=-=-=-e0mlio_( (=-=-) )gbielr_=-=-=-=-=-:Eomn_ nl::::=-=-=-=-=-=_icafee( (-=-=) )elro_er-=-=-=-=-=-::::ln ii::.=-=-=-=-=-=-6digtng( |-=| )fvpm_e0=-=-=-=-=-=-:..:e ee_::t-=-=-=-=-=-=mlio_lel( |=-| )bielr__i-=-=-=-=-=-=::::bm b:::l-=-......=-=-cafeenfv( || )lro_er6d=-=-......=-=cII:i f_::::=-=......-=-=igtngbiel( || )fvpm_e0ml-=-=......-=-:_nr_u nr:::.-=-......=-=-io_lelro_-=-| |=-=ielr__ica=-=-......=-=::::o0 e::::-=-=......-=-=-=-=-=-=-=-=| |-=-=-=-=-=-=-=-=......-=-=::::o b:::o=-=-=-=-=-=-=-=-=-=elr__ic| |nfvpm_e-=-=-=-=-=-=-=-=-=-::::c _s::m-=-=-=-=-=-=-=-=-=-o_er6di| |bielr__=-=-=-=-=-=-=-=-=-=:e_:m p____:-=-=-=-=-=-=-=-=-=m_e0mli| |lro_er6-=-=-=-=-=-=-=-=-=:eomgg eotu2:=-=-=-=-=-=-=-=-=-r((icaf-=-=vpm_))m=-=-=-=-=-=-=-=-=-REmpfn 6::::-=-=-=-=-=-=-=-=-=( (gt=-=-el) )-=-=-=-=-=-=-=-=-=_::Ri ec..:_-=-=-=-=-=-=-=-=( o_-=-=o_ )-=-=-=-=-=-=-=-=::::rn m:::e=-=-=-=-=-=-=-=-( =-=-=-=- )=-=-=-=-=-=-=-=-.:..e grIILe=-=-=-=-=-=-=-=( -=-=-=-= )-=-=-=-=-=-=-=-o2::le eeGEf:=-=-=-=-=-=-=-( (=-=-=-=-) )=-=-=-=-=-=-=-0IDL:c l:::.:=-=-=-=-=-=-=-( =-=-=-=-=- )=-=-=-=-=-=-=-:::mN_ cb....:=-=-=-=-=-=-=-=(=-=-=-=-=-=-)-=-=-=-=-=-=-=-:...:il _v::::2=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-:_:::ii oR0IDL:-=-=-=-=-=-=-=-=-= -=-=-=-=-=-=-=-=-=:lAGO:_ rpAMNr::=-=-=-=-=-=-=-=- =-=-=-=-=-=-=-=-Ea:nRIln eo:::::::-=-=-=-=-=-=-= -=-=-=-=-=-=-=:::::t_ie od::.:::b:-=-=-=-=-=- =-=-=-=-=-=:.:::::r_v an:::::ee::::-=-=-= -=-=-=:_R0IDl::::ln i_t:no2oi:::::ti :::...:bEAMn:bii cbii::::::::: ::::.::CIin_f pbmm_teo0 mi_oi_tle

Intelligence from
incomplete
measurements.

We build mathematically grounded AI that sees beyond limited data—from biomedical images to embodied perception.

Explore research ↓

Department of Robotics & Mechatronics Engineering | DGIST

Recruiting

Now Recruiting Graduate Students & Postdocs

BIFIL is looking for motivated M.S. and Ph.D. students, as well as postdoctoral researchers, who want to work on biomedical imaging and frontier AI — inverse problems, generative models, and trustworthy quantification from theory to clinic.

I am building BIFIL from the ground up — early members will help shape its research direction and culture, with my full attention and close mentorship from day one. If you are interested, please email me with your CV.

Position Inquiries ↗

02 / Team & Leadership

Seung Kwan Kang, Ph.D.

Assistant Professor · DGIST

A medical imaging and AI researcher working across inverse problems, AI, and clinical translation. Previously AI Research Director at Brightonix Imaging and Investment Director at Samsung Ventures.

skang@dgist.ac.kr ↗

Education

2014 – 2021
Ph.D. in Biomedical SciencesSeoul National University College of Medicine
2010 – 2014
B.S. in Electrical and Computer EngineeringSeoul National University

Professional Experience

2026 – present
Assistant ProfessorDGIST · Department of Robotics & Mechatronics Engineering
2026
Investment DirectorSamsung Ventures
2021 – 2026
Director, AI/AlgorithmBrightonix Imaging Inc.

Honors & Awards

2025
NMMI Outstanding Research AwardKorean Society of Nuclear Medicine
2025
Outstanding Paper Award in Science and TechnologyKorean Federation of Science and Technology Societies
2021
Outstanding Graduate AwardSeoul National University
2020
Young Investigator Award (1st place)Korean Society of Nuclear Medicine
2017 – 2020
Education GrantsIEEE NSS/MIC · IEEE NPSS (×6)

Ph.D. Students

AI pioneer. Research interests:

M.S. Students

AI pioneer. Research interests:

Undergraduate Students

AI pioneer. Research interests:

03 / Research

One idea,
three directions.

Infer what matters when measurements are noisy, sparse, or incomplete.

Inverse problems & generative learning

We design principled algorithms that recover structure—and uncertainty—from incomplete measurements.

Inverse problems
Diffusion & flow
Posterior sampling

Biomedical imaging & quantification

We turn PET, SPECT, CT, and MRI into reliable quantitative evidence for research and clinical decisions.

PET / SPECT
Quantitative AI
Clinical translation

Embodied perception & robotics

We extend measurement-aware intelligence to systems that must perceive, reason, and act in the physical world.

Robot learning
VLA
Perception–action

04 / Publications

Theory that
travels.

In Preparation

In preparation

First Author

D3M: data-consistent denoising diffusion model for real-world image denoising problems

SK Kang, JS Lee

2026

Physics in Medicine & Biology· Accepted

Corresponding

Time-aware frame-aligned diffusion for late dynamic PET frame prediction from early frames

YJ Seol, E Kim, SK Kang*, JS Lee

PET denoising

Journal of Nuclear Medicine· doi:10.2967/jnumed.125.271172

Featured

Development and first-in-human imaging results of PHAROS: a versatile high-resolution TOF/DOI PET scanner for brain, breast, and extremity imaging

GB Ko, KY Kim, SK Kang, J Lee, R Jung, DJ Kwak, J-W Son, SA Shin, JS Lee

PET Instrumentation

2025

Journal of Nuclear Medicine· 66(11): 1827–1833

Co-First Author

Artificial intelligence–powered quantification of flortaucipir PET for detecting tau pathology

HB Yoo†, SK Kang†, SA Shin, D Kim, H Choi, YK Kim, D Yi, MS Byun, DY Lee, JS Lee

Tau PET & Quantitative AI

2024

British Journal of Radiology· 97(1155): 632–639

Co-First Author

Sixty-four-fold data reduction of chest radiographs using a super-resolution convolutional neural network

JG Nam†, SK Kang†, H Choi, W Hong, J Park, JM Goo, JS Lee, CM Park

Super-Resolution & Data Reduction

Nuclear Medicine and Molecular Imaging· 58(4): 246–254

First Author

Clinical performance evaluation of an artificial intelligence-powered amyloid brain PET quantification method

SK Kang, M Heo, JY Chung, D Kim, SA Shin, H Choi, A Chong, J-M Ha, H Kim, JS Lee

Clinical AI & Amyloid PET

Nuclear Medicine and Molecular Imaging· 58(6): 354–363

First Author

Accurate automated quantification of dopamine transporter PET without MRI using deep learning-based spatial normalization

SK Kang, D Kim, SA Shin, YK Kim, H Choi, JS Lee

MRI-Free Quantification for DAT PET

Journal of Nuclear Medicine· 65(10): 1645–1651

Corresponding

Improving 18F-FDG PET quantification through a spatial normalization method

D Kim, SK Kang*, SA Shin, H Choi, JS Lee

FDG PET Quantification

Physics in Medicine & Biology· 69(21): 215020

Enhancing bone scan image quality: an improved self-supervised denoising approach

SY Yie, SK Kang, J Gil, D Hwang, H Choi, YK Kim, JC Paeng, JS Lee

Self-Supervised Denoising

2023

Nuclear Medicine and Molecular Imaging· 57(2): 86–93

Automatic lung cancer segmentation in [18F]FDG PET/CT using a two-stage deep learning approach

J Park, SK Kang, D Hwang, H Choi, S Ha, JM Seo, JS Eo, JS Lee

Lesion Segmentation

Journal of Nuclear Medicine· 64(4): 659–666

First Author

Fast and accurate amyloid brain PET quantification without MRI using deep neural networks

SK Kang, D Kim, SA Shin, YK Kim, H Choi, JS Lee

Deep Learning-based Quantification

Frontiers in Nuclear Medicine· 3: 1326549

Editorial

Editorial: Rising stars in PET and SPECT: 2022

D Izquierdo-Garcia, SK Kang

Editorial

2022

European Journal of Nuclear Medicine and Molecular Imaging· 49(6): 1833–1842

Comparison of deep learning-based emission-only attenuation correction methods for positron emission tomography

D Hwang, SK Kang, KY Kim, H Choi, JS Lee

Attenuation Correction

European Journal of Nuclear Medicine and Molecular Imaging· 49(9): 3061–3072

Image-level trajectory inference of tau pathology using variational autoencoder for flortaucipir PET (for Alzheimer's Disease Neuroimaging Initiative)

J Hong, SK Kang, I Alberts, J Lu, R Sznitman, JS Lee, A Rominger, H Choi, K Shi

Generative VAE & Trajectory

2021

Molecular Psychiatry· 26(7): 3476–3488

Dopamine dysregulation in psychotic relapse after antipsychotic discontinuation: an [18F]DOPA and [11C]raclopride PET study in first-episode psychosis

S Kim, SH Shin, B Santangelo, M Veronese, SK Kang, JS Lee, GJ Cheon, W Lee, JS Kwon, OD Howes

Clinical Neuroimaging

Scientific Reports· 11(1): 1673

Co-First Author

Deep learning-based 3D inpainting of brain MR images

SK Kang†, SA Shin, S Seo, MS Byun, DY Lee, YK Kim, DS Lee, JS Lee

3D Image Inpainting

NeuroImage· 232: 117890

First Author

Translating amyloid PET of different radiotracers by a deep generative model for interchangeability (for Alzheimer's Disease Neuroimaging Initiative)

SK Kang, H Choi, JS Lee

Deep Generative Translation

Physics in Medicine & Biology· 66(9): 095010

First Author

Anatomy-guided PET reconstruction using the ℓ1-Bowsher prior

SK Kang, JS Lee

Inverse problem & PET Reconstruction

Physics in Medicine & Biology· 66(11): 115009

Data-driven respiratory phase-matched PET attenuation correction without CT

D Hwang, SK Kang, KY Kim, H Choi, S Seo, JS Lee

CT-Free Attenuation Correction

Biomedical Engineering Letters· 11(3): 263–271

Co-First Author

Synthetic CT generation from weakly paired MR images using cycle-consistent GAN for MR-guided radiotherapy

SK Kang†, HJ An, H Jin, J Kim, EK Chie, JM Park, JS Lee

CycleGAN & Cross-Modality

Electronics· 10(13): 1529

First Author

Noise2Noise improved by trainable wavelet coefficients for PET denoising

SK Kang, SY Yie, JS Lee

Self-Supervised Denoising

Electronics· 10(15): 1836

Accurate transmission-less attenuation correction method for amyloid-β brain PET using deep neural network

BH Choi, D Hwang, SK Kang, KY Kim, H Choi, S Seo, JS Lee

Deep Attenuation Correction

2020

Physica Medica· 72: 60–72

First Author

Robust nonlinear parameter estimation in tracer kinetic analysis using infinity norm regularization and particle swarm optimization

SK Kang, S Seo, C-H Lee, MJ Kim, SJ Kim, JS Lee

Tracer Kinetic Modeling

Nuclear Medicine and Molecular Imaging· 54: 299–304

Self-supervised PET denoising

SY Yie, SK Kang, D Hwang, JS Lee

Denoising

2019

Journal of Nuclear Medicine· 60(8): 1183–1189

Generation of PET attenuation map for whole-body time-of-flight 18F-FDG PET/MRI using a deep neural network trained with simultaneously reconstructed activity and attenuation maps

D Hwang, SK Kang, KY Kim, S Seo, JC Paeng, DS Lee, JS Lee

Deep Attenuation Correction

2018

Physics in Medicine & Biology· 63(11): 115015

Novel inter-crystal scattering event identification method for PET detectors

MS Lee, SK Kang, JS Lee

Detector Physics

Journal of Nuclear Medicine· 59(10): 1624–1629

Improving the accuracy of simultaneously reconstructed activity and attenuation maps using deep learning

D Hwang, KY Kim, SK Kang, S Seo, JC Paeng, DS Lee, JS Lee

Reconstruction & Deep Learning

Human Brain Mapping· 39(9): 3769–3778

First Author

Adaptive template generation for amyloid PET using a deep learning approach

SK Kang, S Seo, SA Shin, MS Byun, DY Lee, YK Kim, DS Lee, JS Lee

Adaptive Deep Templates

Physics in Medicine & Biology· 63(14): 145011

Computed tomography super-resolution using deep convolutional neural network

J Park, D Hwang, KY Kim, SK Kang, YK Kim, JS Lee

CT Super-Resolution

2017

Anticancer Research· 37(3): 1139–1148

Co-First Author

Relationship between Ktrans and K1 with simultaneous versus separate MR/PET in rabbits with VX2 tumors

KH Lee†, SK Kang†, JM Goo, JS Lee, GJ Cheon, S Seo, EJ Hwang

Kinetic Parameters & PET/MR

05 / Projects

Ongoing
research.

O1DGIST

DGIST Start-up Fund

Institutional start-up funding supporting the establishment of BIFIL’s research program in computational biomedical imaging and frontier AI.

Completed Research

P1Korea Health Industry Development Institute (KHIDI)

Brain Aid on Your Smartphone (BAYS)

Development of a healthy aging service using a brain age and cognitive reserve-based foundation model and a digital therapeutics platform.

P2Korea Technology and Information Promotion Agency for SMEs (TIPA)

AI Solution for Radioisotope Therapy Dosimetry

Development of an AI solution and platform for internal radioisotope therapy dosimetry and treatment planning.

P3Korea Health Industry Development Institute (KHIDI)

Longitudinal Neuroimaging Analysis for Dementia

Development of an AI-based longitudinal neuroimaging analysis platform for monitoring dementia progression and treatment response.

Intelligence fromincompletemeasurements.

Now Recruiting Graduate Students & Postdocs

Seung Kwan Kang, Ph.D.

One idea,three directions.

Inverse problems & generative learning

Biomedical imaging & quantification

Embodied perception & robotics

Theory thattravels.

D3M: data-consistent denoising diffusion model for real-world image denoising problems

Time-aware frame-aligned diffusion for late dynamic PET frame prediction from early frames

Development and first-in-human imaging results of PHAROS: a versatile high-resolution TOF/DOI PET scanner for brain, breast, and extremity imaging

Artificial intelligence–powered quantification of flortaucipir PET for detecting tau pathology

Sixty-four-fold data reduction of chest radiographs using a super-resolution convolutional neural network

Clinical performance evaluation of an artificial intelligence-powered amyloid brain PET quantification method

Accurate automated quantification of dopamine transporter PET without MRI using deep learning-based spatial normalization

Improving 18F-FDG PET quantification through a spatial normalization method

Enhancing bone scan image quality: an improved self-supervised denoising approach

Automatic lung cancer segmentation in [18F]FDG PET/CT using a two-stage deep learning approach

Fast and accurate amyloid brain PET quantification without MRI using deep neural networks

Editorial: Rising stars in PET and SPECT: 2022

Comparison of deep learning-based emission-only attenuation correction methods for positron emission tomography

Image-level trajectory inference of tau pathology using variational autoencoder for flortaucipir PET (for Alzheimer's Disease Neuroimaging Initiative)

Dopamine dysregulation in psychotic relapse after antipsychotic discontinuation: an [18F]DOPA and [11C]raclopride PET study in first-episode psychosis

Deep learning-based 3D inpainting of brain MR images

Translating amyloid PET of different radiotracers by a deep generative model for interchangeability (for Alzheimer's Disease Neuroimaging Initiative)

Anatomy-guided PET reconstruction using the ℓ1-Bowsher prior

Data-driven respiratory phase-matched PET attenuation correction without CT

Synthetic CT generation from weakly paired MR images using cycle-consistent GAN for MR-guided radiotherapy

Noise2Noise improved by trainable wavelet coefficients for PET denoising

Accurate transmission-less attenuation correction method for amyloid-β brain PET using deep neural network

Robust nonlinear parameter estimation in tracer kinetic analysis using infinity norm regularization and particle swarm optimization

Self-supervised PET denoising

Generation of PET attenuation map for whole-body time-of-flight 18F-FDG PET/MRI using a deep neural network trained with simultaneously reconstructed activity and attenuation maps

Novel inter-crystal scattering event identification method for PET detectors

Improving the accuracy of simultaneously reconstructed activity and attenuation maps using deep learning

Adaptive template generation for amyloid PET using a deep learning approach

Computed tomography super-resolution using deep convolutional neural network

Relationship between Ktrans and K1 with simultaneous versus separate MR/PET in rabbits with VX2 tumors

Ongoingresearch.

DGIST Start-up Fund

Brain Aid on Your Smartphone (BAYS)

AI Solution for Radioisotope Therapy Dosimetry

Longitudinal Neuroimaging Analysis for Dementia

Intelligence from
incomplete
measurements.

One idea,
three directions.

Theory that
travels.

Ongoing
research.