Sean I Young, PhD

Page 1 Back to the top

Date Prepared

Oct 15, 2025

Personal Details

Sean I. Young, PhD

Assistant Professor of Radiology

Harvard Medical School

Office Address

MGH/HST Martinos Center

149 13th Street

Boston, MA 02129

United States of America

P: +1 617 758 9783

E: siyoung@mgh.harvard.edu

Education

03/2003–11/2007 BEng Software Engineering University of Auckland

Awarded 05/2008 Auckland, New Zealand

07/2010–10/2011 MEngSc Computer Science University of New South Wales

Awarded 08/2011 Sydney, NSW, Australia

03/2012–02/2018 PhD Electrical Engineering University of New South Wales

Awarded 07/2018 (Advisor: David Taubman) Sydney, NSW, Australia

Postdoctoral Training

04/2019–10/2020 Research Department of Electrical Engineering Stanford University

Scholar (Advisor: Bernd Girod) Stanford, CA

10/2020–05/2023 Research Department of Radiology Harvard Medical School

Fellow (Advisor: Bruce Fischl) Boston, MA

Faculty Academic Appointments

11/2025–Present Assistant Professor Department of Radiology Harvard Medical School

05/2023–10/2025 Instructor Department of Radiology Harvard Medical School

Appointments at Affiliated Institutions

10/2021–Present Investigator MGH/HST Martinos Center MGH

Radiology for Biomedical Imaging

10/2021–Present Research Computer Science and Artificial MIT

Affiliate Intelligence Lab (CSAIL)

Page 2 Back to the top

Other Professional Positions

12/2007–05/2009 Programmer Software Engineering Intergen Limited

Auckland, New Zealand

07/2009–06/2010 Programmer Software Engineering Datacom Systems Limited

Auckland, New Zealand

01/2011–12/2011 Tester Software Engineering Free Software Foundation

Boston, MA

07/2016–01/2017 Research Intern Electrical Engineering InterDigital Communications

San Diego, CA

Teaching Positions

2015–2016 TA (3 hrs biweekly Multimedia Signal Processing University of New South Wales

(two semesters) for 16 wks) (Fourth-year EE course) Sydney, NSW, Australia

2017 Tutor (1.5 hrs Multimedia Signal Processing University of New South Wales

weekly for 6 wks) (Fourth-year EE course) Sydney, NSW, Australia

2022–2023 Lecturer (2.5 hrs Artificial Intelligence Kaplan Australia and

(four trimesters) weekly for 48 wks) and Machine Learning New Zealand (Online)

Major Administrative Leadership Positions

Local

2023–2024 MGH CML (MGH Martinos Center for Machine Learning) Host

International

2023 Auckland–Boston Workshop on AI-Driven Medical Imaging Organizer

Professional Societies

2013–Present Institute of Electrical and Electronics Engineers (IEEE) Member

2013–Present IEEE Signal Processing Society Member

2018–Present Australian Pattern Recognition Society (APRS) Member

Report of Funded and Unfunded Projects

Past

2023–2024 Principal K99AG081493 (Ultra-precision Clinical NIH

(Awarded) Investigator Imaging of AD Using Deep Learning)

Current

2025–2028 Principal R00AG081493 (Ultra-precision Clinical NIH

(Awarded) Investigator Imaging of AD Using Deep Learning)

Editorial Activities

Page 3 Back to the top

Ad hoc Reviewer

2017–2018 IEEE Int. Conf. on Image Processing Reviewer

2020–Present IEEE Transactions on Multimedia Reviewer

2020–Present IEEE Transactions on Image Processing Reviewer

2021–Present Nature Scientific Reports Reviewer

2022–Present Frontiers in Artificial Intelligence Reviewer

2022–Present Information Processing in Medical Imaging Reviewer

2023–Present Medical Image Computing and Computer Assisted Intervention Reviewer

2023–Present IEEE/CVF Conf. on Computer Vision and Pattern Recognition Reviewer

2024–Present International Conference on Machine Learning Reviewer

Honors and Prizes

2011 University of New South Wales CSE Performance Award Award

2012–2015 University of New South Wales APA Faculty Award Scholarship

2012–2015 University of New South Wales APA Award (Ph.D.) Scholarship

2018 Australian Pattern Recognition Society Best Paper Award Award

Report of Local, National and International Invited Presentations and Teaching

Local Invited Presentations

No presentations below were sponsored by 3

parties/outside entities

2016 Graph-based regularization for signal processing UC San Diego

(Group seminar) San Diego, CA

2018 Solving vision problems via filtering University of New South Wales

(Group seminar) Sydney, NSW, Australia

2018 “Solving vision problems via filtering” Stanford University

(Group seminar) Stanford, CA

2021 “Transform quantization for CNN compression” ARM Research

(Group seminar) Boston, MA

2021 “Non-line-of-sight Surface Reconstruction” CSAIL, MIT

(Group seminar) Cambridge, MA

2021 “Supervision by denoising for medical image segmentation” McGovern Institute

(Group seminar) Cambridge, MA

2021 “Supervision by denoising for medical image segmentation” CSAIL, MIT

(Group Seminar) Cambridge, MA

2022 “Non-line-of-sight Surface Reconstruction Using the D-LCT” Boston University

(Group seminar) Boston, MA

Page 4 Back to the top

2023 “Fully Convolutional Slice-to-volume Reconstruction McGovern Institute

(Group seminar) Cambridge, MA

National Invited Presentations

No presentations below were sponsored by 3

parties/outside entities

2018 “Fast optical flow extraction from compressed video” DICTA Conference

(Conference presentation) Canberra, ACT, Australia

2020 “NLOS surface reconstruction using the directional LCT” CVPR Conference

(Conference presentation) Online

2021 “Transform quantization for CNN compression” University of Southern California

(Group seminar) Los Angeles, CA

2021 “Transform quantization for CNN compression” UC San Diego

(Group seminar) San Diego, CA

2021 “Transform quantization for CNN compression” Stanford Compression Workshop

(Group seminar) Stanford, CA

2022 “Supervision by Denoising” NIST

(Group seminar) Bethesda, MD

2022 “Supervision by Denoising” University of Maryland

(Group seminar) Bethesda, MD

2023 “Supervision by Denoising” ICCP Conference

(Conference presentation) Madison, WI

2024 “Fully convolutional slice-to-volume Harvard FAS

reconstruction for single-stack MRI” (Group seminar) Cambridge, MA, USA

International Invited Presentations

No presentations below were sponsored by 3

parties/outside entities

2021 “Supervision by denoising for medical segmentation” University College London

(Group seminar) London, UK

2021 “Transform quantization for CNN compression” Facebook AI Toronto

(Group seminar) Toronto, ON, Canada

2021 “Transform quantization for CNN compression” Simon Fraser University

(Group seminar) Vancouver, BC, Canada

2021 “Transform quantization for CNN compression” York University

(Group seminar) Toronto, ON, Canada

2023 “Fully convolutional slice-to-volume Auckland Bioengineering Institute

reconstruction for single-stack MRI” (Group seminar) Auckland, New Zealand

2024 “Fully convolutional slice-to-volume CVPR Conference

reconstruction for single-stack MRI” (Group seminar) Seattle, WA, USA

Page 5 Back to the top

2025 “Radio: Rate–Distortion Optimization for Large Language ICML Conference

Model Compression Vancouver, BC, Canada

Report of Technological and Other Scientific Innovations

2016 “Higher-order motion models and graduated motion estimation” US Patent (PCT)

(Patent) (W02018209067A1)

Report of Formal Teaching and Training

Research Fellows

2023–2024 Xiangrui Zhang (postdoctoral fellow, MGH/HMS, 1 hour/week for 40 weeks)

2023–2024 Ting Gong (postdoctoral fellow, MGH/HMS, 1 hour/week for 10 weeks)

2024–2025 Chiara Mauri (postdoctoral fellow, MGH/HMS, 1 hour/week for 10 weeks)

2024–2025 Xiaoling Hu (postdoctoral fellow, MGH/HMS, 1 hour/week for 10 weeks)

2023–2025 Karthik Gopinath (postdoctoral fellow, MGH/HMS, 1 hour/week for 20 weeks)

2023–2024 Kathleen Larson (postdoctoral fellow, MGH/HMS, 1 hour/week for 40 weeks)

Other Mentored Trainees and Faculty

2025 Yingcheng Liu (PhD student, MIT, 1 hour/week for 5 weeks)

2024–Current Margherita Firenze (PhD student, MIT, 1 hour/week)

Research Supervisory and Training Responsibilities

2024–Current Martinos Center LCN K99/R00 Grant Writing Workshop

Report of Scholarship

Peer-reviewed Scholarship: Research Investigations

1. S. I. Young, A. Naman, D. Taubman. “COGL: Coefficient graph Laplacians for optimized JPEG

image decoding,” IEEE Trans. Image Process. 2019; 28:343–355

2. S. I. Young, A. Naman, B. Girod, D. Taubman. “Solving vision problems via filtering,” Proc. ICCV,

2019

3. S. I. Young, A. Naman, D. Taubman. “Graph Laplacian regularization for robust optical flow

estimation,” IEEE Trans. Image Process. 2020; 29:3970–83

4. S. I. Young, B. Girod, D. Taubman. “Gaussian lifting for fast bilateral and nonlocal means filtering,”

IEEE Trans. Image Process. 2020; 29:6082–95

5. S. I. Young, B. Girod, D. Taubman. “Fast optical flow extraction from compressed video,” IEEE

Trans. Image Process. 2020; 29:6409–21

6. S. Kim, A. Sharma, Y. Liu, S. I . Young. “Rethinking Satellite Data Merging: From averaging to

SNR optimization,” IEEE Trans. Geosci. Remote Sens. 2021

7. S. I. Young, W. Zhe, D. Taubman, B. Girod. “Transform quantization for CNN compression,” IEEE

Trans. Pattern Anal. Mach. Intell. 2021

Page 6 Back to the top

8. S. M. Abulnaga, S. I. Young, K. Hobgood, E. Pan, C. J. Wang, P. E. Grant, E. A. Turk, P. Golland.

“Automatic Segmentation of the Placenta in BOLD MRI Time Series,” Lecture Notes in Computer

Science, vol 13575, Springer

9. S. I. Young, A. V. Dalca, E. Ferrante, P. Golland, C. A. Metzler, J. E. Iglesias, B. Fischl. “Supervision

by denoising,” IEEE Trans. Pattern Anal. Mach. Intell. 2023, 1–1

10. S. I. Young, Y. Balbastre, B. Fischl, P. Golland, J. E. Iglesias. “A framework for interpretability in

machine learning for medical imaging,” IEEE Access, 53277–53292, 2024

11. S. M. Abulnaga, N. Dey, S. I. Young, E. Pan, K. I. Hobgood, Clinton J. Wang, P. E. Grant, E. A.

Turk, P. Golland. “Shape-aware Segmentation of the Placenta in BOLD Fetal MRI Time Series,”

Mach. Learn. Biomed. Imaging. 2023, 2:527–546 .

Peer-reviewed Scholarship: Full-length Conference Proceedings

1. S. I. Young, D. Taubman. “Rate–distortion optimized optical flow estimation,” Proc. IEEE ICIP

2015

2. S. I. Young, R. Mathew, D. Taubman. “Optimizing block-coded motion parameters with block-

partition graphs,” Proc. IEEE ICIP 2016

3. S. I. Young, D. B. Lindell, B. Girod, D. Taubman, G. Wetzstein. “Non-line-of-sight surface

reconstruction using the directional light-cone transform,” Proc. CVPR, 2020

4. S. I. Young, Y. Balbastre, A. V. Dalca, W. M. Wells, J. E. Iglesias, B. Fischl. “SuperWarp: Supervised

Learning and Warping on U-Net for Invariant Subvoxel-Precise Registration” Medical Image

Computing and Computer Assisted Intervention Workshops. 2022, 103–115.

5. M. G. French, G. D. Maso Talou, T. P. Babarenda Gamage, M. P. Nash, J. E. Iglesias, Y. Balbastre,

S. I. Young. “Learning strategies for Breast MR image registration under diffeomorphic constraints,”

Medical Image Computing and Computer Assisted Intervention Workshops. 2023

6. M. Chan, S. I. Young, C. A. Metzler, “SUD2 for semi-supervised computational imaging,” NeurIPS

Workshops. 2023

7. Pablo Blasco Fernandez, Karthik Gopinath, John Williams-Ramirez, Rogeny Herisse, Lucas J

Deden-Binder, Dina Zemlyanker, Theressa Connors, Liana Kozanno, Derek Oakley, Bradley

Hyman, Sean I Young, Juan Eugenio Iglesias. “Pseudo-rendering for Resolution and Topology-

Invariant Cortical Parcellation,” Proc. MLMI. 2024, 74–84.

8. S. I. Young, Y. Balbastre, B. Fischl, P. Golland, J. E. Iglesias. “Fully Convolutional Slice-to-Volume

Reconstruction for Single-Stack MRI,” Proc. CVPR, 2024.

9. S. I. Young, “Radio: Rate–Distortion Optimization for large language model compression,” Proc.

ICML, 2025.

Non-Peer Reviewed Scholarship in Print and Other Media

1. S. I. Young, “Reference-Free 3D Reconstruction of Brain Dissection Photographs with Machine

Learning,” Preprint.

2. S. I. Young, “Radioforms are rate–distortion optimal transforms for large language model

compression,” Preprint.

Page 7 Back to the top

Abstracts, Poster Presentations and Exhibits Presented at Professional Meetings

1. M. Firenze, S. I. Young, C. Wang, E. Adalsteinsson, K. Im, H. Yun, P. E. Grant, P. Golland, “Rapid

Volumetric Reconstruction of the Fetal Brain MRI,” AI Cures Mass General Brigham Symposium.

Thesis

1. S. I. Young. “Graph regularization for inverse problems in imaging,” PhD Thesis, School of Electrical

Engineering and Telecommunications, University of New South Wales, Sydney, NSW, Australia,

June 2018

Narrative Report

I am a computer scientist and electrical engineer whose career has centered on building the algorithmic and

theoretical foundations for next-generation neuroimaging. My academic journey began in computer science

and software engineering, where I developed a deep interest in the mathematics of inverse problems and

optimization. During my doctoral training at the University of New South Wales, I focused on graph-

regularized methods for solving imaging inverse problems, leading to new approaches for high-precision

motion estimation and image reconstruction. These efforts culminated in several publications in IEEE

Transactions on Image Processing and ICCV, advancing the theoretical understanding of regularization

and filtering for vision problems.

At Stanford University, I broadened my focus to computational imaging and machine learning, developing

algorithms for neural network compression and non-line-of-sight (NLOS) surface reconstruction. These

projects unified information theory and optimization to make high-dimensional inference problems both

tractable and interpretable. The insights from these algorithms have directly informed my later work in

neuroimaging—particularly my ongoing effort to make MRI acquisition and reconstruction more efficient

and robust.

Since joining Harvard Medical School and the Martinos Center, I have brought this computational

perspective to medical imaging. My work bridges signal processing, deep learning, and neuroscience

through the design of efficient, data-driven models for 3D brain reconstruction and image registration. For

example, my recent work on fully convolutional slice-to-volume reconstruction enables accurate 3D MRI

synthesis from minimal input data, allowing for improved imaging of populations that cannot tolerate long

scan times. This line of work, along with my contributions to denoising-based supervision and rate–

distortion–optimized neural networks, exemplifies my broader goal: to unify the principles of compression

and reconstruction across artificial and biological systems.

Through these projects, I aim to build a research program in computational radiology—integrating ideas

from machine learning, information theory, and neuroimaging to push toward clinically deployable AI

systems. My current R00 project, Ultra-precision Clinical Imaging of Alzheimer’s Disease Using Deep

Learning, represents the first step toward this vision, translating algorithmic innovation into measurable

clinical benefit. As I transition to the Assistant Professor rank, I plan to expand this interdisciplinary agenda

through collaborations that span engineering, neuroscience, and medicine, training the next generation of

researchers who will define the computational future of radiology.

Select Publications

Page 8 Back to the top

Date Prepared

Oct 15, 2025

Personal Details

Sean I. Young, PhD

Instructor of Radiology

Harvard Medical School

Select Publications

1. Fully Convolutional Slice-to-Volume Reconstruction for Single-Stack MRI (SVR)

Citation: Sean I. Young, Yaël Balbastre, Bruce Fischl, Polina Golland, Juan Eugenio Iglesias. “Fully

Convolutional Slice-to-Volume Reconstruction for Single-Stack MRI.” Proceedings of the

IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), 2024, pp. 11535–

11545. doi: 10.1109/CVPR52733.2024.01096.

Summary: I conceived and led the development of a fully convolutional framework that reconstructs

3D MRI volumes from a single 2D slice stack by directly predicting per-slice motion, then splatting

and interpolating to form the volume. I designed the Splat–Slice U-Net, created the training data

generation pipeline, and conducted the full experimental evaluation on adult and fetal brain MRI.

Compared to optimization-based SVR and recent learning approaches, our method achieves

markedly lower motion error and enables accurate volumetric imaging without multiple stacks—

substantially reducing scan time and patient burden. The approach reframes SVR as supervised

motion estimation, unlocking real-time reconstruction and expanding feasibility for time-sensitive

applications such as fetal fMRI. This contribution is already shaping subsequent advances in

transformer- and implicit-representation-based SVR methods.

2. Supervision by Denoising (SUD)

Citation: Sean I. Young, Adrian V. Dalca, Enzo Ferrante, Polina Golland, Christopher A. Metzler,

Bruce Fischl, Juan Eugenio Iglesias. “Supervision by Denoising.” IEEE Transactions on Pattern

Analysis and Machine Intelligence (TPAMI), 47(9):7279–7291, September 2025. doi:

10.1109/TPAMI.2023.3299789. PMID: 37505997.

Summary: I originated the SUD framework and led the technical development, implementing an

alternating scheme in which a network’s denoised outputs serve as pseudo-labels to supervise itself,

thereby reducing dependence on scarce pixel-accurate annotations. I unified temporal ensembling

and spatial denoising into a single spatio-temporal framework, specified the loss formulation, and

carried out comprehensive evaluations on 3D anatomical brain reconstruction and 2D cortical

parcellation. SUD improved reconstruction quality and generalization over supervised-only and

ensembling baselines while remaining domain-agnostic, eliminating the need for handcrafted task-

specific regularizers. This work broadens the scope of semi- and self-supervised learning in medical

imaging and is increasingly cited in both methodological and applied contexts. The NIH-funded

nature of this project (K99 AG081493) underscores its translational and clinical relevance.

Page 9 Back to the top

3. Radio: Rate–Distortion Optimization for Large Language Model Compression

Citation: Sean I. Young. “Radio: Rate–Distortion Optimization for Large Language Model

Compression.” Proceedings of the International Conference on Machine Learning (ICML), 2025,

PMLR 267: 14532–14544.

Summary: I developed Radio, a principled framework that formulates neural-network compression

as a rate–distortion optimization problem, unifying information theory and deep learning. Radio

minimizes a Lagrangian objective that jointly penalizes quantization bitrate and task-specific

distortion, yielding analytically optimal precision allocation across layers and parameters. By deriving

closed-form distortion models and implementing an efficient calibration pipeline, I demonstrated that

Radio attains 4-bit quantization of LLaMA- and GPT-class models with negligible performance loss

(< 0.3 % in perplexity). Within my R00 project, this framework forms the computational backbone

for speech-based diagnosis of Alzheimer’s disease and related dementias (ADRD)—enabling

deployment of clinically focused large language models on low-power hospital or mobile devices.