About

Garry Gold is a professor associated with Radiology and Musculoskeletal Imaging at Stanford University. His academic focus involves medical imaging, specifically related to musculoskeletal systems. As part of the faculty at the Stanford Program in Human Biology, he contributes to the interdisciplinary environment that integrates research and education in human biology and medicine. His work is recognized within the university's broader medical and biological research community, contributing to advancements in imaging techniques and their applications in human health.

Research topics

• Medicine
• Artificial Intelligence
• Computer Science
• Pathology
• Radiology
• Internal medicine
• Nuclear medicine
• Anatomy
• Mathematics
• Biology
• Chemistry
• Statistics
• Intensive care medicine

Selected publications

• Advances in cartilage imaging techniques

  Nature Reviews Rheumatology · 2026-02-03 · 2 citations

  article
  PublisherDOI

• Smartphone video–based knee extension moments during chair rise relate to MRI measures of muscle function

  medRxiv · 2026-03-09

  articleOpen access

  Background: Preserving muscle function is essential for maintaining independence during aging, but muscle force-generating capacity is not commonly measured clinically due to a lack of accessible, sensitive tools. Magnetic resonance imaging (MRI) provides gold-standard measures of muscle volume and microstructure, which reflect force-generating capacity, while dynamometry quantifies peak joint moments during voluntary contraction. Both modalities are time-consuming and costly, so clinical and large-scale studies often rely on low-fidelity measures such as the time to complete the five times sit-to-stand test (5xSTS). OpenCap, a tool for quantifying musculoskeletal dynamics from smartphone videos, may provide an accessible and more informative approach to assessing muscle function. We evaluated whether OpenCap-derived knee extension moments during chair rise relate to MRI-based measures of quadriceps muscle volume and microstructure, using dynamometry as a comparator. Methods: Nineteen healthy adults of various ages (63.2% female, 57.8 ± 15.4 y, 30-78 y) underwent quadriceps MRI, dynamometry, and 5xSTS time with concurrent OpenCap data collection. Using MRI, we computed quadriceps volume and radial diffusivity (a measure related to fiber size). We standardized these features and summed to create a composite MRI score, reflecting muscle quantity and quality. We estimated peak knee extension moment using OpenCap during chair rise and via both isometric and isokinetic dynamometry. We compared OpenCap kinematics (torso angle) and dynamics (knee moment), 5xSTS time, and dynamometry to MRI measures of muscle function using linear regression; false discovery rate was controlled using the Benjamini-Hochberg procedure. Results: The OpenCap-derived knee extension moment was associated with quadriceps muscle volume (r=0.63, p=0.014) and radial diffusivity (r=0.61, p=0.016). Peak knee extension moments measured by both isometric and isokinetic dynamometry were correlated with muscle volume (r=0.66-0.75, p=0.002-0.009), but not with radial diffusivity (r=0.04-0.52, p=0.054-0.91). Both OpenCap and isokinetic dynamometry showed their strongest associations with the composite MRI score (r=0.77, p=0.002 and r=0.73, p=0.002, respectively). 5xSTS time and a kinematic feature (torso angle) were not associated with any MRI-derived measures (r=-0.16-0.35, p=0.22-0.97). Conclusions: Smartphone video-based joint moments associate with muscle size and microstructure, unlike time or kinematic features. OpenCap offers a scalable assessment of muscle force-generating capacity that can be conducted rapidly without specialized equipment, enabling higher-fidelity assessments of muscle function in the clinic and in large-scale studies where imaging and dynamometry are impractical.

  PublisherOA PDFDOI

• BMI and Varus Malalignment Compound to Define a High-Risk Phenotype for Compartment-Specific Knee Osteoarthritis Progression

  medRxiv · 2026-04-17

  articleOpen access

  Objectives: Knee osteoarthritis (KOA) is a leading cause of disability, yet which patients will experience structural decline remains unclear. Body mass index (BMI) and lower limb alignment are established risk factors for KOA, but their independent and interactive effects on compartment-specific cartilage loss and total knee replacement (TKR) have not been characterized at scale. Methods: We analyzed 5,832 limbs from 3,016 participants in the Osteoarthritis Initiative followed over 7 years. Cartilage thickness in the weight-bearing medial and lateral femur and tibia was quantified, and lower limb alignment was measured using hip-knee-ankle (HKA) angle obtained from full-limb radiographs. Linear mixed-effects models estimated the independent and interactive effects of BMI and lower limb alignment on longitudinal cartilage thinning, and mixed-effects logistic regression modeled TKR risk. Results: BMI and +10° varus, the rate of medial femur cartilage thinning was 243.5% faster than the reference rate. In the lateral compartment, BMI and valgus alignment were independently associated with faster cartilage thinning, with no significant interaction. TKR risk increased exponentially with HKA deviation (odds ratio [OR] = 1.38 per 1°; ~five-fold at 5° malalignment) but was not associated with BMI. Conclusion: BMI and lower limb alignment influence structural KOA progression through compartment-specific pathways. The multiplicative interaction in the medial compartment identifies high BMI combined with varus malalignment as a discrete high-risk phenotype, with implications for clinical risk stratification and disease-modifying intervention design.

  PublisherDOI

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

  IEEE Transactions on Instrumentation and Measurement · 2025-01-01 · 1 citations

  article

  Low-dose fluoroscopy is essential for real-time X-ray visualization, supporting dynamic diagnostic assessments while minimizing harmful radiation exposure to patients. However, low-dose imaging introduces noise that can impair diagnostic accuracy. Although numerous deep learning methods have been developed for noise reduction in medical imaging, the unique challenges of fluoroscopy—such as motion artifacts due to its real-time nature, limited access to clean reference data, and high noise levels—diminish the effectiveness of current deep learning-based denoising techniques, leaving research in this area relatively constrained. To address these challenges, we present three key innovations. First, we propose an unsupervised framework for dynamic, context-aware denoising in fluoroscopy, introducing the multiscale recurrent attention U-Net (MSR2AU-Net) to effectively reduce noise without clean data by directly targeting initial noise. Second, our proposed dual-noise suppression strategy combines a knowledge distillation-based module for uncorrelated noise with a recursive filtering module for correlated noise, enhancing both denoising quality and motion stability. Finally, we design a pixel-wise dynamic object motion cross-fusion matrix combined with an edge-preserving loss function to preserve fine details amidst motion changes. Our model was evaluated on 3500 fluoroscopy images from dynamic phantoms (2400 for training and 1100 for testing) and 350 clinical images from spinal surgery patients, confirming its effectiveness in clinical settings. To further validate the robustness of our approach, we also tested it on the “2016 NIH-AAPM-Mayo Clinic Low Dose CT Grand Challenge” dataset, using 4800 images for training and 1136 for testing, demonstrating efficacy across both fluoroscopy and CT imaging. Quantitative evaluations demonstrated that our approach outperformed existing unsupervised algorithms, achieving a structural similarity index measure (SSIM) of 0.9803 and a peak signal-to-noise ratio (PSNR) of 39.12 dB on the dynamic phantom dataset, and an SSIM of 0.9591 with a PSNR of 36.62 dB on the Mayo CT dataset. These results indicate that our approach surpasses state-of-the-art unsupervised algorithms in both visual quality and quantitative metrics, achieving performance comparable to well-established supervised methods in low-dose fluoroscopy and CT imaging. The source code will be available at <uri xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">https://github.com/sunyoungIT/UDCA-Net.git</uri>.

  PublisherDOI

• NEURAL SHAPE MODEL QUANTIFIES EARLY AND PROGRESSIVE BONE SHAPE CHANGES AFTER ACLR

  Osteoarthritis Imaging · 2025-01-01 · 1 citations

  articleOpen access

  Femoral bone shape scores (B-Score) derived from shape models quantify 3D structural features associated with OA 1,2 . A higher B-Score is indicative of more OA-like bone shape. B-Scores have high sensitivity to quantify OA progression and stratify patients for interventions 1 . Neural Shape Models (NSM) capture non-linear bone shape features and outperform traditional Statistical Shape Models (SSMs) in encoding OA-related shapes 3 . Prior work that used a SSM-based B-Score showed that anterior cruciate ligament reconstructed (ACLR) knees exhibit higher B-Scores than their contralateral knees 2 years post-surgery, reflecting OA-like bone shape features 4 . However, little is known about how femoral bone shape changes immediately following ACLR and how it progresses during the early post-surgical period—a critical window when post-traumatic osteoarthritis (PTOA) may still be most responsive to intervention. To use a Neural Shape Model-based B-Score to quantify femoral shape differences between ACLR and contralateral knees immediately post-surgery (3-weeks) and to detect early PTOA bone shape changes over 30 months. ACLR and contralateral knees of 17 subjects (11M/6F, age=38±10 yrs, BMI=24±2 kg/m 2 ) were scanned at 3 weeks (baseline), 3, 9, 18, and 30 months post-ACLR in a 3T MRI scanner (GE Healthcare, USA) using a qDESS sequence (TE/TR=6/22 ms , flip angle=25°, FOV=160 × 160 mm, bandwidth=31.25 kHz, pixel spacing=0.42 × 0.50 mm, slice thickness=1.5 mm). The femur was automatically segmented, and the B-Score was computed for each subject at all visits using a NSM that was trained on 9,376 femoral segmentations from the baseline DESS images in the OAI dataset 1 . To assess bone shape differences immediately after surgery, we compared B-Scores between the ACLR and contralateral knees at the baseline visit using a linear mixed effects model. To capture longitudinal bone shape changes after surgery, we calculated change in B-Score at each follow-up visit with respect to the baseline visit. We used a linear mixed effects model to assess the effect of knee-type and time post-surgery on B-Scores. Effect sizes [η p 2 is small (0.01), medium (0.06), or large (0.14)] were computed for significant effects (p<0.05). At baseline, the ACLR knee B-Score was significantly lower than the contralateral knee (η p 2 =0.40, p=0.005; Fig. 1A). Longitudinally, ACLR knees showed a significantly greater increase in B-Score than contralateral knees (η p 2 =0.19, p<0.001; Fig 2A). The lower B-Scores in ACLR knees at baseline indicate that the surgical knee had a healthier, less OA-like bone shape than the contralateral knee. Visualization revealed that ACLR knees had a wider intercondylar notch compared to their contralateral knee resulting from notchplasty that were confirmed on surgical notes (Fig. 1B). Since idiopathic OA-like features typically include notch narrowing 2 , the surgically altered geometry, particularly the widened intercondylar notch yields a shape less characteristic of OA, resulting in a lower B-Score. Longitudinally, however, we observe early osteophyte lipping, particularly in the trochlea, intercondylar notch, and medial-posterior condyle—bone shape changes that align with idiopathic OA and likely explain the steep increase in B-Score for ACLR knees over time (Fig. 2B and C). Neural shape modeling characterizes femoral shape changes due to ACLR surgery. Accounting for surgically induced shape changes enables detection of OA-like features as early as 3 months post-ACLR and enhances sensitivity to track these changes longitudinally, potentially serving as a sensitive biomarker for early detection and monitoring of PTOA.

  PublisherDOI

• Comparison of Femoral Neck PET Uptake and MRI Fat Fraction in a Healthy Aging Female Population

  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: Current x-ray-based methods to diagnose osteoporosis preclude a comprehensive evaluation of bone health evaluate bone marrow composition and metabolic activity. Goal(s): Our goal was to evaluate PET SUV and MRI fat fraction in the femoral neck to elucidate age-related differences in a healthy aging female population. Approach: [18F]NaF PET/MRI hip scans using a 3T GE hybrid PET/MRI system was used to assess SUV and fat fraction in 12 healthy females (6 younger, 6 older, 24 hips). Results: Older women had significantly higher fat fraction than younger women. Baseline PET SUV did not differ significantly between groups, but heat maps revealed distinct uptake patterns. Impact: [18F]NaF PET/MRI provides a promising multimodal approach to assess bone health, enabling precise monitoring of osteoporosis progression and treatment effects. By revealing insights into bone marrow composition and metabolic activity, this technique may enhance understanding of age-related bone changes.

  PublisherDOI

• Hamstring muscle architecture and microstructure changes following Nordic hamstring exercise training and detraining

  Journal of sport and health science/Journal of Sport and Health Science · 2025-06-25 · 4 citations

  articleOpen access

  • Nine weeks of Nordic hamstring exercise training substantially increased hamstring muscle volume as well as fiber length and cross-section. • Nordic hamstring exercise induced non-uniform adaptations across the 4 hamstring muscles; the hypertrophy and fiber length elongation were greater in the semitendinosus and biceps femoris short head compared to the semimembranosus and biceps femoris long head. • Upon 3 weeks of detraining, gains in fiber length underwent rapid reversal, but gains in volume and fiber cross-section were mostly retained, which underscores the need for consistent training to sustain all the protective benefits imparted by the exercise. Background While Nordic hamstring exercise (NHE) training has been shown to reduce hamstring strains, the muscle-specific adaptations to NHE across the 4 hamstrings remain unclear. This study investigates architectural and microstructural adaptations of the biceps femoris short head (BFsh), biceps femoris long head (BFlh), semitendinosus (ST), and semimembranosus (SM) in response to an NHE intervention. Methods Eleven subjects completed 9 weeks of supervised NHE training followed by 3 weeks of detraining. Magnetic resonance imaging was performed at pre-training, post-training, and detraining to assess architectural (volume, fiber tract length, and fiber tract angle) and microstructural (axial (AD), mean (MD), radial (RD) diffusivities, and fractional anisotropy (FA)) parameters of the 4 hamstrings. Results NHE training induced significant but non-uniform hamstring muscle hypertrophy (BFsh: 22%, BFlh: 9%, ST: 26%, SM: 6%) and fiber tract length increase (BFsh: 11%, BFlh: 7%, ST: 18%, SM: 10%). AD (5%), MD (4%), and RD (5%) showed significant increases, but fiber tract angle and FA remained unchanged. After detraining, only ST showed a significant reduction (8%) in volume, which remained higher than the pre-training value. While fiber tract lengths returned to baseline, AD, MD, and RD remained higher than pre-training levels for all hamstrings. Conclusion The 9-week NHE training substantially increased hamstring muscle volume with greater hypertrophy in ST and BFsh. Hypertrophy was accompanied by increases in fiber tract lengths and cross-sections (increased RD). After 3 weeks of detraining, fiber tract length gains across all hamstrings declined, emphasizing the importance of sustained training to maintain all the protective adaptations.

  PublisherDOI

• Magnetic resonance imaging provides comparable spinal curvature measurements to computerized tomography

  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: Spinal curvature measurements may be used to better understand scoliosis, lordosis, and low back anatomy. The unknown effect of the type of imaging modality on semi-automatically measured spinal curvature motivated this study, as well as the unknown relationship between image contrast and curvature measurements. Goal(s): To quantify how spinal curvature measurements are impacted by changing imaging modalities, specifically between CT and T1-weighted MR images, and between T1-weighted and T2-weighted MR images. Approach: We compared 20 patients' spinal curvature, calculated four separate ways from their L5-L1 vertebral centroids in each imaging modality. Results: We found no significant differences across modalities in calculating spinal curvature. Impact: This study impacts research done between CT scans and MR images, supporting intra-modality spinal curvature calculations from vertebral centroids calculated semi-automatically. This study also supports the equivalency of spinal curvature calculations from CT scans, T1-weighted and T2-weighted MR images.

  PublisherDOI

• Exploratory analysis of infrapatellar fat pad MRI-based radiomics for detection of knee structure abnormalities in collegiate basketball players and swimmers

  Osteoarthritis Imaging · 2025-10-30

  articleOpen access

  Magnetic resonance imaging (MRI) based radiomic evaluation of the infrapatellar fat pad (IPFP) has been shown to predict knee osteoarthritis. As IPFP abnormalities can arise from sport-related injuries, this study evaluated whether qualitatively identified knee structure abnormalities in young athletes can be detected by IPFP MRI-based radiomics. Bilateral knee MRIs using Dixon fat-water decomposition techniques were obtained from 46 NCAA Division 1 collegiate 46 basketball players (26 male, 20 female) and 21 swimmers (10 male, 11 female). Board-certified musculoskeletal radiologists evaluated anatomic features and patellar height using the modified Noyes score and Caton-Deschamps index.. IPFP volumes were segmented, and fat fraction was computed. Radiomic features were calculated within 2D overlapping patches extracted from the IPFP in the fat and water images separately. Cross-validated logistic regression models were developed using IPFP radiomic features as predictors of an athlete’s sport and the occurrence of cartilage lesions, tendinopathy, or bone abnormalities as observed on MRI. Mann-Whitney U tests evaluated differences in fat fraction between sports and knee structure abnormalities. The area under the receiver operating characteristic (ROC) curve (AUC; maximum 0.79) indicated that IPFP radiomics from fat-only images can differentiate between basketball and swimming athletes. Tendinopathy was identified (AUC = 0.68 ± 0.05) at larger patch sizes. Qualitative radiological assessments of cartilage lesions and bone abnormalities were not distinguished (AUC < 0.57). Fat fraction did not differ across sports or knee structure abnormality (p > 0.49, mean difference < 0.5%). In the absence of inflammatory arthropathy, IPFP radiomics discriminated between sports but not cartilage lesions, tendinopathy, or bone abnormalities, suggesting structural adaptation to sport-specific loading. Abnormal IPFP signal intensity may not present in young athletes without traumatic injury or knee arthroscopy.

  PublisherDOI

• Hamstring Muscle Architecture and Microstructure Changes Following 9-weeks of Nordic Hamstring Exercise Training

  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: Evaluate long-term muscle adaptations across the full volume of all four hamstrings in response to Nordic hamstring exercise (NHE) to enhance injury prevention strategies. Goal(s): Examine how 9-weeks of supervised NHE-training affects architecture (volume, fiber length, angle, and curvature) and microstructure (MD, RD, FA, and T2) of Biceps femoris short head (BFsh), Biceps femoris long head (BFlh), Semitendinosus (ST), and Semimembranosus (SM). Approach: 11 subjects underwent MRI scans (Dixon, DTI, and T2) pre and post 9-weeks NHE-training. Results: NHE-training increased hamstring volume with greater hypertrophy in ST and BFsh muscles. Hypertrophy was accompanied by increases in both length and cross-section of muscle fibers. Impact: This study examines architectural and microstructural adaptations of the hamstrings following 9-weeks of Nordic hamstring exercise training. Findings reveal significant, but non-uniform hypertrophy among hamstrings accompanied by increase in length and size of the muscle fibers, advancing injury prevention strategies.

  PublisherDOI

Recent grants

• Advanced MR Imaging of Early Osteoarthritis

  NIH · $1.5M · 2012–2024

• Weight-Bearing Imaging of the Knee Using C-Arm CT

  NIH · $2.0M · 2015–2021

• NIH Grant P41RR009784

  NIH · $5.4M · 2015

• Designing Novel MR Imaging Tools to Quantify Lower-Limb Exercise Adaptations in Knee Osteoarthritis

  NIH · $9.5M · 2003–2028

• Advanced MR Imaging of Early Osteoarthritis

  NIH · $181k · 2012–2023

Frequent coauthors

• François R. Herrmann

  University of Geneva

  177 shared

• Pantéléimon Giannakopoulos

  University Hospital of Geneva

  173 shared

• F. Eckstein

  Ludwig Boltzmann Institute for Digital Health and Prevention

  133 shared

• Frank W. Roemer

  Universitätsklinikum Erlangen

  132 shared

• Ali Guermazi

  131 shared

• M.D. Crema

  Institut National du Sport, de l'Expertise et de la Performance

  129 shared

• Timothy J. Mosher

  Penn State Milton S. Hershey Medical Center

  128 shared

• Deborah Burstein

  Harvard University

  124 shared

Education

• Ph.D., Human Biology

  Stanford University

  1990

• M.S., Human Biology

  University of California, San Francisco

  1985

• B.A., Human Biology

  University of California, Berkeley

  1980