Also see our publication highlights for a featured list.


  • Robust multishot diffusion-weighted imaging of the abdomen with region-based shot rejection. Magnetic resonance in medicine Lee, P. K., Zhou, X., Hargreaves, B. A. 2024


    Diffusion-weighted (DW) imaging provides a useful clinical contrast, but is susceptible to motion-induced dephasing caused by the application of strong diffusion gradients. Phase navigators are commonly used to resolve shot-to-shot motion-induced phase in multishot reconstructions, but poor phase estimates result in signal dropout and Apparent Diffusion Coefficient (ADC) overestimation. These artifacts are prominent in the abdomen, a region prone to involuntary cardiac and respiratory motion. To improve the robustness of DW imaging in the abdomen, region-based shot rejection schemes that selectively weight regions where the shot-to-shot phase is poorly estimated were evaluated.Spatially varying weights for each shot, reflecting both the accuracy of the estimated phase and the degree of subvoxel dephasing, were estimated from the phase navigator magnitude images. The weighting was integrated into a multishot reconstruction using different formulations and phase navigator resolutions and tested with different phase navigator resolutions in multishot DW-echo Planar Imaging acquisitions of the liver and pancreas, using conventional monopolar and velocity-compensated diffusion encoding. Reconstructed images and ADC estimates were compared qualitatively.The proposed region-based shot rejection reduces banding and signal dropout artifacts caused by physiological motion in the liver and pancreas. Shot rejection allows conventional monopolar diffusion encoding to achieve median ADCs in the pancreas comparable to motion-compensated diffusion encoding, albeit with a greater spread of ADCs.Region-based shot rejection is a linear reconstruction that improves the motion robustness of multi-shot DWI and requires no sequence modifications.

    View details for DOI 10.1002/mrm.30102

    View details for PubMedID 38623901

  • Clinical evaluation of isotropic MAVRIC-SL for symptomatic hip arthroplasties at 3 T MRI. Magnetic resonance imaging Yoon, D., Doyle, Z., Lee, P., Hargreaves, B., Stevens, K. 2024


    3D multi-spectral imaging (MSI) of metal implants necessitates relatively long scan times.We implemented a fast isotropic 3D MSI technique at 3 T and compared its image quality and clinical utility to non-isotropic MSI in the evaluation of hip implants.Two musculoskeletal radiologists scored images from coronal proton density-weighted conventional MAVRIC-SL and an isotropic MAVRIC-SL sequence accelerated with robust-component-analysis on a 3-point scale (3: diagnostic, 2: moderately diagnostic, 1: non-diagnostic) for overall image quality, metal artifact, and visualization around femoral and acetabular components. Grades were compared using a signed Wilcoxon test. Images were evaluated for effusion, synovitis, osteolysis, loosening, pseudotumor, fracture, and gluteal tendon abnormalities. Reformatted axial and sagittal images for both sequences were subsequently generated and compared for image quality with the Wilcoxon test. Whether these reformats increased diagnostic confidence or revealed additional pathology, including findings unrelated to arthroplasty that may contribute to hip pain, was also compared using the McNemar test. Inter-rater agreement was measured by Cohen's kappa.39 symptomatic patients with a total of 59 hip prostheses were imaged (mean age, 70 years ±9, 14 males, 25 females). Comparison scores between coronal images showed no significant difference in image quality, metal artifact, or visualization of the femur and acetabulum. Except for loosening, reviewers identified more positive cases of pathology on the original coronally-acquired isotropic sequence. In comparison of reformatted axial and sagittal images, the isotropic sequence scored significantly (p < 0.01) higher for overall image quality (3.0 vs 2.0) and produced significantly (p < 0.01) more cases of increased diagnostic confidence (42.4% vs 7.6%) or additional diagnoses (50.8% vs 22.9%). Inter-rater agreement was substantial (k = 0.798) for image quality. Mean scan times were 4.2 mins (isotropic) and 7.1 mins (non-isotropic).Compared to the non-isotropic sequence, isotropic 3D MSI was acquired in less time while maintaining diagnostically acceptable image quality. It identified more pathology, including postoperative complications and potential pain-generating pathology unrelated to arthroplasty. This fast isotropic 3D MSI sequence demonstrates promise for improving diagnostic evaluation of symptomatic hip prostheses at 3 T while simultaneously reducing scan time.

    View details for DOI 10.1016/j.mri.2024.04.017

    View details for PubMedID 38621551

  • Stereoscopic calibration for augmented reality visualization in microscopic surgery. International journal of computer assisted radiology and surgery El Chemaly, T., Athayde Neves, C., Leuze, C., Hargreaves, B., H Blevins, N. 2023


    Middle and inner ear procedures target hearing loss, infections, and tumors of the temporal bone and lateral skull base. Despite the advances in surgical techniques, these procedures remain challenging due to limited haptic and visual feedback. Augmented reality (AR) may improve operative safety by allowing the 3D visualization of anatomical structures from preoperative computed tomography (CT) scans on real intraoperative microscope video feed. The purpose of this work was to develop a real-time CT-augmented stereo microscope system using camera calibration and electromagnetic (EM) tracking.A 3D printed and electromagnetically tracked calibration board was used to compute the intrinsic and extrinsic parameters of the surgical stereo microscope. These parameters were used to establish a transformation between the EM tracker coordinate system and the stereo microscope image space such that any tracked 3D point can be projected onto the left and right images of the microscope video stream. This allowed the augmentation of the microscope feed of a 3D printed temporal bone with its corresponding CT-derived virtual model. Finally, the calibration board was also used for evaluating the accuracy of the calibration.We evaluated the accuracy of the system by calculating the registration error (RE) in 2D and 3D in a microsurgical laboratory setting. Our calibration workflow achieved a RE of 0.11 ± 0.06 mm in 2D and 0.98 ± 0.13 mm in 3D. In addition, we overlaid a 3D CT model on the microscope feed of a 3D resin printed model of a segmented temporal bone. The system exhibited small latency and good registration accuracy.We present the calibration of an electromagnetically tracked surgical stereo microscope for augmented reality visualization. The calibration method achieved accuracy within a range suitable for otologic procedures. The AR process introduces enhanced visualization of the surgical field while allowing depth perception.

    View details for DOI 10.1007/s11548-023-02980-5

    View details for PubMedID 37450175

    View details for PubMedCentralID 4634572

  • Noise2Recon: Enabling SNR-robust MRI reconstruction with semi-supervised and self-supervised learning. Magnetic resonance in medicine Desai, A. D., Ozturkler, B. M., Sandino, C. M., Boutin, R., Willis, M., Vasanawala, S., Hargreaves, B. A., Re, C., Pauly, J. M., Chaudhari, A. S. 2023


    PURPOSE: To develop a method for building MRI reconstruction neural networks robust to changes in signal-to-noise ratio (SNR) and trainable with a limited number of fully sampled scans.METHODS: We propose Noise2Recon, a consistency training method for SNR-robust accelerated MRI reconstruction that can use both fully sampled (labeled) and undersampled (unlabeled) scans. Noise2Recon uses unlabeled data by enforcing consistency between model reconstructions of undersampled scans and their noise-augmented counterparts. Noise2Recon was compared to compressed sensing and both supervised and self-supervised deep learning baselines. Experiments were conducted using retrospectively accelerated data from the mridata three-dimensional fast-spin-echo knee and two-dimensional fastMRI brain datasets. All methods were evaluated in label-limited settings and among out-of-distribution (OOD) shifts, including changes in SNR, acceleration factors, and datasets. An extensive ablation study was conducted to characterize the sensitivity of Noise2Recon to hyperparameter choices.RESULTS: In label-limited settings, Noise2Recon achieved better structural similarity, peak signal-to-noise ratio, and normalized-RMS error than all baselines and matched performance of supervised models, which were trained with 14 *

  • [Formula: see text] Field inhomogeneity correction for qDESS [Formula: see text] mapping: application to rapid bilateral knee imaging. Magma (New York, N.Y.) Barbieri, M., Watkins, L. E., Mazzoli, V., Desai, A. D., Rubin, E., Schmidt, A., Gold, G. E., Hargreaves, B. A., Chaudhari, A. S., Kogan, F. 2023


    [Formula: see text] mapping is a powerful tool for studying osteoarthritis (OA) changes and bilateral imaging may be useful in investigating the role of between-knee asymmetry in OA onset and progression. The quantitative double-echo in steady-state (qDESS) can provide fast simultaneous bilateral knee [Formula: see text] and high-resolution morphometry for cartilage and meniscus. The qDESS uses an analytical signal model to compute [Formula: see text] relaxometry maps, which require knowledge of the flip angle (FA). In the presence of [Formula: see text] inhomogeneities, inconsistencies between the nominal and actual FA can affect the accuracy of [Formula: see text] measurements. We propose a pixel-wise [Formula: see text] correction method for qDESS [Formula: see text] mapping exploiting an auxiliary [Formula: see text] map to compute the actual FA used in the model.The technique was validated in a phantom and in vivo with simultaneous bilateral knee imaging. [Formula: see text] measurements of femoral cartilage (FC) of both knees of six healthy participants were repeated longitudinally to investigate the association between [Formula: see text] variation and [Formula: see text].The results showed that applying the [Formula: see text] correction mitigated [Formula: see text] variations that were driven by [Formula: see text] inhomogeneities. Specifically, [Formula: see text] left-right symmetry increased following the [Formula: see text] correction ([Formula: see text] = 0.74 > [Formula: see text] = 0.69). Without the [Formula: see text] correction, [Formula: see text] values showed a linear dependence with [Formula: see text]. The linear coefficient decreased using the [Formula: see text] correction (from 24.3 ± 1.6 ms to 4.1 ± 1.8) and the correlation was not statistically significant after the application of the Bonferroni correction (p value > 0.01).The study showed that [Formula: see text] correction could mitigate variations driven by the sensitivity of the qDESS [Formula: see text] mapping method to [Formula: see text], therefore, increasing the sensitivity to detect real biological changes. The proposed method may improve the robustness of bilateral qDESS [Formula: see text] mapping, allowing for an accurate and more efficient evaluation of OA pathways and pathophysiology through longitudinal and cross-sectional studies.

    View details for DOI 10.1007/s10334-023-01094-y

    View details for PubMedID 37142852

    View details for PubMedCentralID 2268124

  • Improving Data-Efficiency and Robustness of Medical Imaging Segmentation Using Inpainting-Based Self-Supervised Learning. Bioengineering (Basel, Switzerland) Dominic, J., Bhaskhar, N., Desai, A. D., Schmidt, A., Rubin, E., Gunel, B., Gold, G. E., Hargreaves, B. A., Lenchik, L., Boutin, R., Chaudhari, A. S. 2023; 10 (2)


    We systematically evaluate the training methodology and efficacy of two inpainting-based pretext tasks of context prediction and context restoration for medical image segmentation using self-supervised learning (SSL). Multiple versions of self-supervised U-Net models were trained to segment MRI and CT datasets, each using a different combination of design choices and pretext tasks to determine the effect of these design choices on segmentation performance. The optimal design choices were used to train SSL models that were then compared with baseline supervised models for computing clinically-relevant metrics in label-limited scenarios. We observed that SSL pretraining with context restoration using 32 × 32 patches and Poission-disc sampling, transferring only the pretrained encoder weights, and fine-tuning immediately with an initial learning rate of 1 × 10-3 provided the most benefit over supervised learning for MRI and CT tissue segmentation accuracy (p < 0.001). For both datasets and most label-limited scenarios, scaling the size of unlabeled pretraining data resulted in improved segmentation performance. SSL models pretrained with this amount of data outperformed baseline supervised models in the computation of clinically-relevant metrics, especially when the performance of supervised learning was low. Our results demonstrate that SSL pretraining using inpainting-based pretext tasks can help increase the robustness of models in label-limited scenarios and reduce worst-case errors that occur with supervised learning.

    View details for DOI 10.3390/bioengineering10020207

    View details for PubMedID 36829701

  • Development of augmented reality technology for surgical resection accuracy via improved visualization Doan, A. T., Fischer, M. J., Walsh, T. M., Koire, S., Daniel, B. L., Hargreaves, B. A., Black, M. S., Kress, B. C., Peroz, C. SPIE-INT SOC OPTICAL ENGINEERING. 2023

    View details for DOI 10.1117/12.2648463

    View details for Web of Science ID 001170185600040

  • The Impact of Occlusion on Depth Perception at Arm's Length. IEEE transactions on visualization and computer graphics Fischer, M., Rosenberg, J., Leuze, C., Hargreaves, B., Daniel, B. 2023; 29 (11): 4494-4502


    This paper investigates the accuracy of Augmented Reality (AR) technologies, particularly commercially available optical see-through displays, in depicting virtual content inside the human body for surgical planning. Their inherent limitations result in inaccuracies in perceived object positioning. We examine how occlusion, specifically with opaque surfaces, affects perceived depth of virtual objects at arm's length working distances. A custom apparatus with a half-silvered mirror was developed, providing accurate depth cues excluding occlusion, differing from commercial displays. We carried out a study, contrasting our apparatus with a HoloLens 2, involving a depth estimation task under varied surface complexities and illuminations. In addition, we explored the effects of creating a virtual "hole" in the surface. Subjects' depth estimation accuracy and confidence were a ssessed. Results showed more depth estimation variation with HoloLens and significant depth error beneath complex occluding surfaces. However, creating a virtual hole significantly reduced depth errors and increased subjects' confidence, irrespective of accuracy enhancement. These findings have important implications for the design and use of mixed-reality technologies in surgical applications, and industrial applications such as using virtual content to guide maintenance or repair of components hidden beneath the opaque outer surface of equipment. A free copy of this paper and all supplemental materials are available at

    View details for DOI 10.1109/TVCG.2023.3320239

    View details for PubMedID 37782607

  • Multishot Diffusion-Weighted MRI of the Breasts in the Supine vs. Prone Position. Journal of magnetic resonance imaging : JMRI Moran, C. J., Middione, M. J., Mazzoli, V., McKay-Nault, J. A., Guidon, A., Waheed, U., Rosen, E. L., Poplack, S. P., Rosenberg, J., Ennis, D. B., Hargreaves, B. A., Daniel, B. L. 2022


    BACKGROUND: Diffusion-weighted imaging (DWI) may allow for breast cancer screening MRI without a contrast injection. Multishot methods improve prone DWI of the breasts but face different challenges in the supine position.PURPOSE: To establish a multishot DWI (msDWI) protocol for supine breast MRI and to evaluate the performance of supine vs. prone msDWI.STUDY TYPE: Prospective.POPULATION: Protocol optimization: 10 healthy women (ages 22-56), supine vs. prone: 24 healthy women (ages 22-62) and five women (ages 29-61) with breast tumors.FIELD STRENGTH/SEQUENCE: 3-T, protocol optimization msDWI: free-breathing (FB) 2-shots, FB 4-shots, respiratory-triggered (RT) 2-shots, RT 4-shots, supine vs. prone: RT 4-shot msDWI, T2-weighted fast-spin echo.ASSESSMENT: Protocol optimization and supine vs. prone: three observers performed an image quality assessment of sharpness, aliasing, distortion (vs. T2), perceived SNR, and overall image quality (scale of 1-5). Apparent diffusion coefficients (ADCs) in fibroglandular tissue (FGT) and breast tumors were measured.STATISTICAL TESTS: Effect of study variables on dichotomized ratings (4/5 vs. 1/2/3) and FGT ADCs were assessed with mixed-effects logistic regression. Interobserver agreement utilized Gwet's agreement coefficient (AC). Lesion ADCs were assessed by Bland-Altman analysis and concordance correlation (rhoc ). P value <0.05 was considered statistically significant.RESULTS: Protocol optimization: 4-shots significantly improved sharpness and distortion; RT significantly improved sharpness, aliasing, perceived SNR, and overall image quality. FGT ADCs were not significantly different between shots (P=0.812), FB vs. RT (P=0.591), or side (P=0.574). Supine vs. prone: supine images were rated significantly higher for sharpness, aliasing, and overall image quality. FGT ADCs were significantly higher supine; lesion ADCs were highly correlated (rhoc =0.92).DATA CONCLUSION: Based on image quality, supine msDWI outperformed prone msDWI. Lesion ADCs were highly correlated between the two positions, while FGT ADCs were higher in the supine position.EVIDENCE LEVEL: 2.TECHNICAL EFFICACY: Stage 1.

    View details for DOI 10.1002/jmri.28582

    View details for PubMedID 36583628