About

Dr. Peirong Liu is an assistant professor of Electrical and Computer Engineering at Johns Hopkins University, with a joint appointment at the Data Science and Artificial Intelligence (DSAI) Institute. Her research focuses on AI for Healthcare, specifically at the intersection of machine learning, computer vision, and medical image computing. Her work aims to advance foundational theories of learning and representation to establish general, reliable, and accessible frameworks that can understand and support the complex and imperfect nature of real-world healthcare systems. Dr. Liu’s research has been published in top-tier machine learning and computer vision conferences, including CVPR, ICCV, ECCV, ICLR, and NeurIPS, as well as leading medical image computing venues such as IEEE TMI, MICCAI, and IPMI, with multiple works recognized for oral presentations. She earned her PhD in Computer Science from the University of North Carolina at Chapel Hill in 2023 and completed her postdoctoral training at the Athinoula A. Martinos Center for Biomedical Imaging of Harvard Medical School and Massachusetts General Hospital. She has been recognized as a Rising Star in EECS by MIT and as a Rising Star in Data Science by UCSD, UChicago, and Stanford.

Research topics

• Chemistry
• Materials science
• Computer Science
• Nanotechnology
• Radiology
• Pathology
• Biophysics
• Biology
• Internal medicine
• Photochemistry
• Biochemistry
• Medicine

Selected publications

• 265 Surgical Revascularization Resolves Asymmetry in Cortico-Cerebellar Vascular Reactivity and Resting-State Connectivity in Anterior Circulation Moyamoya

  Neurosurgery · 2026-03-26

  article
  PublisherDOI

• Blood‐oxygenation-level-dependent (BOLD) MRI responses to CO2 and O2 inhalation in brain gliomas

  Magnetic Resonance Imaging · 2025-02-28 · 1 citations

  article
  PublisherDOI

• Cerebrovascular Reactivity Development in Infants during the First Two Years of Life Using Resting-State BOLD Functional MRI

  Proceedings on CD-ROM - International Society for Magnetic Resonance in Medicine. Scientific Meeting and Exhibition/Proceedings of the International Society for Magnetic Resonance in Medicine, Scientific Meeting and Exhibition · 2025-09-16

  article

  Motivation: There is no systematic research focused on the development of cerebrovascular reactivity (CVR) in infants. Goal(s): This study aims to investigate the early development of CVR with resting-state BOLD analysis. Approach: We mapped CVR from BCP dataset and fitted developmental trajectory from 1 to 24months. The relationship between CVR and cognitive abilities as well as its association with neurotransmitter receptors/transporters was also investigated. Results: Multiple brain regions showed significant age effect of CVR. CVR at the left superior frontal gyrus was associated with expressive language. GABAaR, mGluR5 and VAChT were found to be related to CVR development as well. Impact: We demonstrated the feasibility of utilizing resting-state BOLD to evaluate the development of CVR in infants, revealing the potential neurotransmitter receptors/transporters system involved. This approach offers a novel perspective on assessing cerebrovascular health in infants.

  PublisherDOI

• Iterative immunogen optimization to focus immune responses on a conserved, subdominant influenza hemagglutinin epitope

  bioRxiv (Cold Spring Harbor Laboratory) · 2025-11-06

  preprintOpen access

  Abstract Designing effective vaccination strategies against genetically diverse viruses, such as HIV or influenza, is hindered by the ability of these pathogens to mutate and readily evade immune responses. While these viruses contain conserved regions that cannot easily mutate without affecting viral fitness, these sites are subdominant, meaning that they are not the primary target of antibodies elicited by natural infection or current vaccines. Here, we integrated recent advances in machine learning methods for protein structure prediction and design to engineer protein immunogens that focus immune responses on a conserved, subdominant HA epitope targeted by broadly cross-reactive influenza antibodies. Iterating between computation-guided optimization and in vivo analyses generated immunogens that faithfully displayed the target epitope and redirected humoral responses toward it upon vaccination. These results provide a blueprint for applying recent protein engineering approaches to immunogen design and may inform the development of broadly protective influenza vaccines.

  PublisherOA PDFDOI

• The dependence of CO2 cerebrovascular reactivity (CVR) on caffeine

  Imaging Neuroscience · 2025-01-01 · 1 citations

  articleOpen access

  Abstract Cerebrovascular reactivity (CVR) represents an important marker of brain vascular health, particularly in the context of small and large vessel diseases. However, an undesired feature of this measure is that there exist large variations in CVR values across individuals, which is mainly attributed to physiological factors. Here, we test the hypothesis that caffeine, a widely consumed neurostimulant, has a significant effect on CVR measured with MRI. Sixteen young healthy participants were enrolled and categorized into caffeine-naive (N = 8) and caffeine-habituated (N = 8) groups based on their caffeine consumption habits. CVR was assessed via CO2 inhalation using two different MRI methods, phase-contrast (PC) cerebral blood flow (CBF), and T2*-EPI Blood-Oxygenation-Level-Dependent (BOLD)-MRI. Each participant underwent two MRI sessions, one before and the other after an oral administration of 200 mg of caffeine. Additionally, venous oxygenation (Yv) was measured using T2-Relaxation-Under-Spin-Tagging (TRUST) MRI. For basal physiological parameters, a significant caffeine-induced CBF decrease was observed in both naive (p = 0.002) and habituated (p < 0.001) groups. The caffeine-naive group exhibited a 31.2 ± 14.1% reduction in basal CBF, whereas the caffeine-habituated group showed a 16.7 ± 5.0% reduction, revealing significant differences between groups (p = 0.04). A similar observation was seen in basal Yv, with caffeine-naive participants showing a greater (p = 0.02) reduction (21.5 ± 8.9%) than the habituated participants (7.6 ± 10.1%). CBF-CVR decreased significantly in both groups: from 4.5 ± 0.9 to 3.0 ± 0.9 %CBF/mmHg of CO2 (33.3 ± 14.1%, p < 0.001) in the caffeine-naive group, and from 5.1 ± 1.5 to 3.7 ± 1.3 %CBF/mmHg of CO2 (27.3 ± 16.0%, p = 0.009) in the caffeine-habituated group. No significant differences were observed between groups in terms of the extent of CVR reduction (p = 0.23). BOLD-CVR showed modest reduction after caffeine administration, from 0.17 ± 0.04 %/mmHg to 0.15 ± 0.05 %/mmHg (14.1 ± 16.8%, p = 0.02). There was no difference between the participant groups in terms of BOLD-CVR reduction following caffeine consumption. This study suggests that investigations using CVR as a disease marker may benefit from accounting for the caffeine consumption and/or its blood concentration in the participants.

  PublisherOA PDFDOI

• Relationship between brain oxygen metabolism and cognition in older adults

  Proceedings on CD-ROM - International Society for Magnetic Resonance in Medicine. Scientific Meeting and Exhibition/Proceedings of the International Society for Magnetic Resonance in Medicine, Scientific Meeting and Exhibition · 2025-09-16

  article

  Motivation: A reduction in brain metabolism is a hallmark of neurodegeneration in Alzheimer's disease (AD), according to recent criteria for staging AD. However, alterations in brain metabolism during the preclinical phase of AD are not fully understood. Goal(s): To investigate the association between brain metabolism and cognitive performance at a very early stage of dementia, when cognition is still normal. Approach: Using non-invasive MRI, we measured cerebral metabolic rate of oxygen (CMRO2) and examined its association with cognitive test scores. Results: CMRO2 was lower in participants with MCI/dementia compared to those cognitively unimpaired. However, among the unimpaired group, CMRO2 was inversely associated with cognition. Impact: The present work suggests that, at the very early stage of Alzheimer's disease, higher brain oxygen metabolism is a sign of lower cognitive function, reflecting brain's hypermetabolism/hyperactivity during preclinical neurodegeneration.

  PublisherDOI

• Automatic Quality Control for Resting‐State BOLD‐Based Cerebrovascular Reactivity Mapping

  NMR in Biomedicine · 2025-12-03

  article

  ABSTRACT Cerebrovascular reactivity (CVR) mapping based on resting‐state BOLD fMRI can be widely available for research of vascular health not only in clinical studies but also in open databases. However, as it utilizes spontaneous CO 2 fluctuations of blood as endogenous stimuli, resting‐state CVR may be prone to low SNR and reproducibility if the CO 2 fluctuation of an individual is small. The automatic identification of such poor‐quality CVR datasets is crucial for large‐scale research. Thus, in this work, we developed an automatic quality control algorithm for resting‐state CVR mapping. Utilizing a total of 51 resting‐state CVR maps acquired with three scanning protocols in each healthy participant, quality control parameters reflecting common characteristics of poor‐quality CVR, including pooled variance of different tissue types, proportion of negative voxels in gray matter, and the sensitivity of the BOLD signal to CVR, were extracted and then combined into one comprehensive quality evaluation index (QEI). We further evaluated its performance by leave‐one‐out cross‐validation and correlation analyses with test–retest reproducibility. Leave‐one‐out cross‐validation showed that QEI was significantly correlated with the reference standard of quality evaluation in all left‐out cases ( r = 0.766). Correlation analyses with test–retest reproducibility revealed significant positive correlations between the worse QEI and similarity index of CVR maps from two tests ( r = 0.809, 0.890, 0.396, and 0.654 for data from four open databases). The proposed QEI performed not only in good agreement with visual inspection but can also adapt in resting‐state CVR from multiple age groups and scanning protocols, paving the way for the clinical applications of resting‐state CVR mapping technology.

  PublisherDOI

• ASL perfusion MRI in neonates less than one week of age: arterial suppression and vector-projection-based CBF estimation

  Proceedings on CD-ROM - International Society for Magnetic Resonance in Medicine. Scientific Meeting and Exhibition/Proceedings of the International Society for Magnetic Resonance in Medicine, Scientific Meeting and Exhibition · 2025-09-16

  article

  Motivation: Arterial-spin labeling (ASL) MRI in newborns less than one week of age is arguably the most challenging application of ASL. Goal(s): To develop an ASL scheme that provides high-fidelity perfusion maps while minimizing noise and biases. Approach: We developed a projection-based data processing method to reduced CBF overestimation, and verified its efficacy using simulation, phantom and neonatal ASL data. We further identified optimal arterial saturation to suppress pulsation noise while preserving SNR. Results: Projection-based complex-subtraction method can mitigate perfusion overestimation bias. Arterial saturation with a 15 cm/s cutoff velocity effectively reduced large-vessel signal fluctuations without affecting tissue perfusion signal. Impact: Using the proposed ASL acquisition and processing method, high-fidelity perfusion maps can be obtained from early neonates in less than 4 minutes. This technique holds potentials for studying neonatal brain diseases involving perfusion abnormalities.

  PublisherDOI

• CVR Transfer Functional Analysis and multi-PLD pCASL MRI to examine vascular dynamics during subacute stroke recovery

  Proceedings on CD-ROM - International Society for Magnetic Resonance in Medicine. Scientific Meeting and Exhibition/Proceedings of the International Society for Magnetic Resonance in Medicine, Scientific Meeting and Exhibition · 2025-09-16

  article

  Motivation: The subacute phase of stroke recovery is crucial for rehabilitation, but the vascular dynamics of repair and angiogenesis remain poorly understood Goal(s): To examine the capability of physiological MRI to delineate the vascular dynamics during subacute stroke recovery Approach: CVR transfer function analysis (TFA) and 3D-pCASL (2200 and 2800 msec post-labeling delays) were collected from a subacute stroke patient longitudinally. Results: The TFA results show that the lesion and surrounding areas stabilize during recovery, with the vascular steal phenomenon occurring in the affected hemisphere. The multi-delay pCASL technique effectively shows a collateral flow that supports recovery and angiogenesis. Impact: Given that stroke has a vascular etiology, characterizing the vascular remodeling and associated dynamics, especially during the subacute phase is crucial to developing effective stroke rehabilitation strategies. Advanced physiological MRI techniques designed for this purpose can significantly influence clinical practice.

  PublisherDOI

• High‐Fidelity <scp>MRI</scp> Assessment of Cerebral Perfusion in Healthy Neonates Less Than 1 Week of Age

  Journal of Magnetic Resonance Imaging · 2025-08-09

  articleOpen access
  PublisherOA PDFDOI

Recent grants

• Calibration of fMRI in Emotional Aging

  NIH · $465k · 2019–2022

• NIH Grant R21NS100006

  NIH · $450k · 2021

• Assessment of brain oxygen consumption in neonates using MRI

  NIH · $2.4M · 2018–2026

• Gas-free cerebrovascular reactivity (CVR) MRI in vascular cognitive impairment

  NIH · $436k · 2020–2021

• NIH Grant R21NS08563401A1

  NIH · $26k · 2014

Frequent coauthors

• Hanzhang Lu

  Johns Hopkins University

  506 shared

• Shin‐Lei Peng

  China Medical University

  88 shared

• Zixuan Lin

  72 shared

• Dengrong Jiang

  71 shared

• Yang Li

  63 shared

• Binu P. Thomas

  62 shared

• Peter C.M. van Zijl

  Johns Hopkins University

  57 shared

• Jay J. Pillai

  Mayo Clinic

  54 shared

Awards & honors

• Rising Star in EECS by MIT
• Rising Star in Data Science by UCSD, UChicago and Stanford