About

Gigi Galiana, PhD, is an Associate Professor of Radiology and Biomedical Imaging at Yale School of Medicine. Her research focuses on medical imaging, particularly in the areas of mammography, radiology, and magnetic resonance imaging. She has contributed to advancements in imaging techniques such as 3D phase contrast using balanced steady-state free precession for improved 4D flow at clinical field strengths, and diffusion-weighted imaging of prostate cancer using nonlinear gradient coils. Her work involves developing innovative imaging methods to enhance the detection and characterization of various medical conditions, with a particular emphasis on prostate cancer and cardiovascular imaging. Dr. Galiana holds a PhD from Princeton University, obtained in 2008, and is actively involved in research at Yale Biomedical Imaging Institute, Yale Cancer Center, and Yale Ventures. Her research includes the development of nonlinear gradient coils, high-resolution diffusion imaging, and the standardization of prostate MRI imaging protocols. She collaborates frequently with other researchers in the field, contributing to peer-reviewed publications that advance the understanding and application of magnetic resonance imaging technologies in clinical settings.

Research topics

• Medicine
• Medical physics
• Radiology
• Computer Science
• Biology
• Chemistry
• Physics
• Cancer research
• Nuclear magnetic resonance
• Internal medicine
• Nuclear medicine
• Biochemistry
• Acoustics

Selected publications

• Subject-Specific Low-Field MRI Synthesis via a Neural Operator

  ArXiv.org · 2026-03-26

  articleOpen access

  Low-field (LF) magnetic resonance imaging (MRI) improves accessibility and reduces costs but generally has lower signal-to-noise ratios and degraded contrast compared to high field (HF) MRI, limiting its clinical utility. Simulating LF MRI from HF MRI enables virtual evaluation of novel imaging devices and development of LF algorithms. Existing low field simulators rely on noise injection and smoothing, which fail to capture the contrast degradation seen in LF acquisitions. To this end, we introduce an end-to-end LF-MRI synthesis framework that learns HF to LF image degradation directly from a small number of paired HF-LF MRIs. Specifically, we introduce a novel HF to LF coordinate-image decoupled neural operator (H2LO) to model the underlying degradation process, and tailor it to capture high-frequency noise textures and image structure. Experimental results in T1w and T2w MRI demonstrate that H2LO produces more faithful simulated low-field images than existing parameterized noise synthesis models and popular image-to-image translation models. Furthermore, it improves performance in downstream image enhancement tasks, showcasing its potential to enhance LF MRI diagnostic capabilities.

  PublisherOA PDF

• <scp>3D</scp> Phase Contrast Using Balanced Steady‐State Free Precession ( <scp>PC</scp> ‐ <scp>SSFP</scp> ) for Improved <scp>4D</scp> Flow at Clinical Field Strengths

  Magnetic Resonance in Medicine · 2026-01-26

  articleOpen access

  PURPOSE: 4D flow MRI provides comprehensive evaluation of cardiovascular flow. One major limitation of intracardiac 4D flow is poor blood-myocardial contrast. 2D phase contrast using balanced steady state free precession (PC-SSFP) methods have been demonstrated to provide accurate velocity, with enhanced contrast and SNR. In this work, we extended our 2D PC-SSFP to 4D flow at clinical field strengths, and tested it for diastolic evaluation. METHODS: The 4D flow sequence with four-point encoding was modified to have 0th and 1st moment gradient nulling over each TR to achieve bSSFP contrast. Pixel-wise velocities were validated in a flow phantom at 3 T. Mitral inflow peak velocity (E, A, e') and stroke volume (SV) were compared in 14 scan (13 healthy subjects at 3 T, with one subject scanned again at 1.5 T) with standard 2D and 4D flow GRE methods. RESULTS: In phantom study, 4D flow bSSFP strongly agreed with GRE, with r > 0.9 in all three directions. Significantly improved SNR (42.4 ± 24.7 vs. 16.9 ± 8.5) and blood-tissue CNR (9.7 ± 3.3 vs. 2.3 ± 1.5) were found in vivo. 4D flow bSSFP measured comparable E (limits of agreement 1.3 ± 14.6 cm/s, r = 0.88), A (0.3 ± 11.3 cm/s, r = 0.95), e' (1.3 ± 3.4 cm/s, r = 0.71), and SV (-2.6 ± 9.7 mL, r = 0.91) vs. GRE approach, and showed similar agreement with 2D methods (r = 0.64-0.91). A study at 1.5 T suggested its potential applicability at lower field strength, with reduced susceptibility to off-resonance artifacts. CONCLUSION: Our 4D flow bSSFP method is feasible, achieving improved SNR, CNR, and accurately measuring mitral velocity and volume at clinical field strengths.

  PublisherOA PDFDOI

• Strong‐Gradient Diffusion‐Weighted Imaging of Prostate Cancer Using an Inside‐Out Nonlinear Gradient Coil

  Magnetic Resonance in Medicine · 2026-01-08

  articleOpen accessSenior author

  ABSTRACT Purpose To demonstrate improved image quality and lesion conspicuity in prostate diffusion‐weighted imaging (DWI) using an inside‐out nonlinear gradient coil that provides locally strong gradients (200–500 500 mT/m) at typical prostate positions. Theory and Methods Before applying the nonlinear gradient coil to DWI with Echo Planar Imaging (EPI) readout, we investigated geometric distortion and eddy currents, and proposed necessary corrections. We then developed two DWI protocols ( b max = 1000 and 3000 s/mm 2 ) with minimized echo time (TE) and tested them on volunteers and patients. We validated apparent diffusion coefficient (ADC) maps from the nonlinear gradient acquisition against the reference (linear gradients only). We quantified improvements in signal‐to‐noise ratio (SNR), lesion contrast‐to‐noise ratio (CNR), and lesion‐to‐normal‐tissue contrast ratio in the compartmental map of restricted diffusion. Results Corrections effectively reduced nonlinear‐gradient DWI artifacts. ADC maps from linear‐ and nonlinear‐gradient‐encoded studies agreed well, with a normalized root‐mean‐square‐error of ∼10%, a common level of ADC variation. TE was significantly reduced from 57 to 42–47 ms for moderate b ‐values (≤ 1000 s/mm 2 ) and from 72 to 42–54 ms for high b ‐values (≤ 3000 s/mm 2 ). Consequently, SNR increased by 3%–38% (median 16%, p &lt; 0.01) and 7%–38% (median 26%, p &lt; 0.01), respectively. Lesion CNR improved by a median of 133% at b = 2000 s/mm 2 and 217% at b = 3000 s/mm 2 . The restricted diffusion component in lesions was more conspicuous at short TE, with a median 23% increase in lesion‐to‐normal‐tissue contrast ratio ( p = 0.02). Conclusion The inside‐out nonlinear gradient coil enhances prostate DWI.

  PublisherDOI

• Subject-Specific Low-Field MRI Synthesis via a Neural Operator

  arXiv (Cornell University) · 2026-03-26

  preprintOpen access

  Low-field (LF) magnetic resonance imaging (MRI) improves accessibility and reduces costs but generally has lower signal-to-noise ratios and degraded contrast compared to high field (HF) MRI, limiting its clinical utility. Simulating LF MRI from HF MRI enables virtual evaluation of novel imaging devices and development of LF algorithms. Existing low field simulators rely on noise injection and smoothing, which fail to capture the contrast degradation seen in LF acquisitions. To this end, we introduce an end-to-end LF-MRI synthesis framework that learns HF to LF image degradation directly from a small number of paired HF-LF MRIs. Specifically, we introduce a novel HF to LF coordinate-image decoupled neural operator (H2LO) to model the underlying degradation process, and tailor it to capture high-frequency noise textures and image structure. Experimental results in T1w and T2w MRI demonstrate that H2LO produces more faithful simulated low-field images than existing parameterized noise synthesis models and popular image-to-image translation models. Furthermore, it improves performance in downstream image enhancement tasks, showcasing its potential to enhance LF MRI diagnostic capabilities.

  PublisherDOI

• Patients’ pain experience in claudication: biopsychosocial correlates of frequency, intensity, sleep interference, and chronic pain

  Journal of Pain · 2026-05-01

  article
  PublisherDOI

• Microstructural Imaging with a Nonlinear Gradient: Pushing the Limit of Short Diffusion Time

  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

  articleSenior author

  Motivation: Locally high gradient strength from a nonlinear gradient paves a new way for microstructure MRI at short diffusion times well below 10ms, and the geometry of the nonlinear gradient suits skeletal muscle of lower extremities. Goal(s): To derive radial diffusivity of calf muscle at short diffusion times and derive the microstructure. Approach: We designed a diffusion encoding scheme integrating linear and nonlinear gradients, and developed Radial Diffusivity Imaging. We estimate surface-to-volume ratio from the radial diffusivity at short diffusion times. Results: With diffusion times as short as 4ms, the time-dependent radial diffusivity shows expected square-root time dependence, enabling quantification of calf muscle microstructure. Impact: With the high strength of nonlinear gradient, this study reveals time-dependent diffusivity of the calf muscle at short diffusion times, and sheds light on microstructure of the calf muscle non-invasively.

  PublisherDOI

• Prostate Diffusion-Weighted Imaging with an Inside-Out Nonlinear Gradient Coil

  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

  articleSenior author

  Motivation: Widely used to diagnose prostate cancer, diffusion-weighted imaging (DWI) is limited by gradient strength. As an alternative to costly whole-body strong-gradient scanners, we proposed an insertable inside-out nonlinear gradient coil tailored for the prostate. Goal(s): Achieve the required diffusion weighting in the prostate with high signal-to-noise ratio (SNR) and contrast-to-noise ratio (CNR). Approach: We developed DWI methods for the nonlinear gradient, including eddy current correction and design of diffusion encoding scheme. Results: The nonlinear gradient reduces echo time (42ms vs. 57ms or 72ms), enhances prostate SNR, and shows promise for high lesion CNR. Impact: The proposed inside-out nonlinear gradient coil provides easy access to strong gradient for high-SNR and high-CNR prostate DWI, which can be flexibly installed for clinical scanners and facilitates triage, diagnosis, and surveillance of prostate cancer.

  PublisherDOI

• Enhanced standardization of clinical T2-weighted prostate images: e-CAMP with T2 prior

  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

  articleSenior author

  Motivation: Standardized T2-weighted images are valuable for prostate cancer healthcare. Although e-CAMP, a standardization algorithm for clinical T2-weighted images, has been proven feasible in brains, challenges remain for prostates. Goal(s): Developing an enhanced algorithm of e-CAMP to remove aliasing and ringing artifacts and avoid underestimation. Approach: Based on the insight that T2-weighted images and T2 maps are correlated, a T2 prior image is created from T2-weighted images to guide the reconstruction of T2 in e-CAMP. Results: While T2 prior approximates well, e-CAMP is closer to the ground truth, indicating unique contributions from both parts. Preliminary results on fastMRI prostate dataset are also encouraging. Impact: Sensitivity of machine learning to scanner- and/or protocol- variability can be reduced by estimating T2 maps from T2w. Here we present the feasibility of this approach for prostate imaging, where machine learning has shown great promise.

  PublisherDOI

• Decoupling Low Field Array Coil Using High Impedance Preamplifier

  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: Phased array coils can benefit low field MRI via higher SNR or parallel imaging. However, isolation preamplifiers at very low frequency are difficult to source, and the bandwidth of LC resonance circuits at very low frequency can be narrow. Goal(s): To investigate a different approach to array coil decoupling at low field. Approach: Instead of using an LC network with low Z preamplifier, a high Z preamplifier with RF transformer can be directly used to decouple array coils. Results: More than 20dB coil decoupling over a bandwidth of 1MHz was achieved around a readout frequency of 1MHz. Impact: A simple and efficient method to decoupling array coil at low field is investigated in this work. Without the need of complicated matching and customize-built isolation preamplifier, broadband and sufficient decoupling on array element can be achieved.

  PublisherDOI

• 4D flow with a bSSFP approach

  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: Conventional 4D flow may find application in diastolic dysfunction evaluations, but is limited by low contrast. Goal(s): Our goal was to extend our previously established 2D PC-SSFP to 4D flow for more robust mitral flow measurements, simultaneously providing E, A, e' and possibly tricuspid regurgitation in a single scan. Approach: We investigated our method in both flow phantoms and healthy volunteers, and compared with 2D CMR and spoiled GRE based 4D flow. Results: Our proposed 4D flow PC-SSFP measured similar flow and maximum velocities, while achieving improved myocardium-blood contrast and higher SNR. Impact: A 4D flow PC-SSFP was developed for more robust diastolic dysfunction evaluations, which estimated mitral inflow and valve velocity in a single scan, and demonstrated agreement with 2D measurements and spoiled GRE based 4D flow, yet improved image quality.

  PublisherDOI

Recent grants

• NIH Grant F31EB005979

  NIH · $133k · 2008

• Practical Nonlinear Gradient Encoding for Enhanced Accelerated Imaging

  NIH · $1.5M · 2017–2023

• Narrow linewidth MRS to detect breast cancer

  NIH · $659k · 2012–2017

• Distinguishing Lipid Subtypes by Amplifying Contrast from J-Coupling

  NIH · $461k · 2018–2022

Frequent coauthors

• R. Todd Constable

  Yale University

  37 shared

• Rosa T. Branca

  18 shared

• Nahla Elsaid

  18 shared

• Dana C. Peters

  Yale University

  16 shared

• Zhehong Zhang

  15 shared

• Leo Tam

  Nvidia (United States)

  14 shared

• Yonghyun Ha

  13 shared

• Warren S. Warren

  13 shared