Muyinatu Bell
· John C. Malone Associate ProfessorVerifiedJohns Hopkins University · Electrical and Computer Engineering
Active 1972–2026
About
Muyinatu Bell is recognized internationally for her pioneering work in medical imaging technology, specifically ultrasound and photoacoustic imaging, photoacoustic-guided surgery, robot-assisted imaging, machine learning for image formation, and other cutting-edge techniques created to significantly advance healthcare interventions and diagnosis. She is the John C. Malone Associate Professor in the Department of Electrical and Computer Engineering at Johns Hopkins University, with a joint appointment in the Department of Biomedical Engineering and a secondary appointment in the Department of Computer Science. Bell founded and directs the Photoacoustic and Ultrasonic Systems Engineering (PULSE) Lab at Johns Hopkins University, where she developed and patented the world’s first short-lag spatial coherence (SLSC) beamformer for ultrasound data. Her team has created innovations such as software that trains computers to remove background noise from ultrasounds, providing clearer images for interventional radiologists, which earned her the Trailblazer Award from the National Institute of Biomedical Imaging and Bioengineering in 2018. Her research also explores photoacoustic imaging systems for surgical guidance, leading to her recognition as one of MIT Technology Review’s “35 Innovators Under 35” in 2016, Maryland’s Outstanding Young Engineer in 2019, and an Alfred P. Sloan Research Fellow in 2019. Her lab’s work on optical fiber-based photoacoustic sensing of blood vessels and nerves aims to eliminate surgical complications caused by accidental injury. Additionally, she introduced the concept of teleoperated photoacoustic-guided surgery using robotic systems, combining imaging and minimally invasive surgery to improve accuracy. Bell’s extensive professional affiliations include collaborations with Johns Hopkins’ Carnegie Center for Surgical Innovation, Laboratory for Computational Sensing Robotics, Malone Center for Engineering in Healthcare, and the Johns Hopkins School of Medicine Department of Oncology. She has received numerous awards, fellowships, and honors, including election as a fellow of Optica and SPIE, and recognition by the National Academy of Engineering as a Senior Member. Bell has published over 60 scientific articles, holds multiple patents, and actively participates in professional societies and editorial roles, contributing significantly to the fields of biomedical imaging and engineering.
Research topics
- Computer Science
- Artificial Intelligence
- Physics
- Acoustics
- Optics
- Computer vision
- Telecommunications
- Medical physics
- Computational biology
- Biology
- Medicine
- Biomedical engineering
Selected publications
Investigating photoacoustic imaging with augmented reality for surgical guidance applications
2026-03-09
articleSenior authorIEEE Transactions on Ultrasonics · 2026-01-12
articleOpen accessSenior authorIn medical diagnostics, ultrasound strain elastography offers a non-invasive method to assess tissue stiffness, aiding in distinguishing tumors, cysts, and benign lesions from healthy tissue. A strain elastography framework employs a speckle-tracking technique to estimate the displacement between pre- and post-compression radiofrequency (RF) frames, which is spatially differentiated to calculate the strain field. Second-order ultrasound elastography (SOUL) is a recently published speckle-tracking algorithm that optimizes a regularized cost function comprising data fidelity and regularization terms. However, the success of SOUL to produce high-quality strain images depends on the manual selection of regularization parameters, making it user-dependent and limiting its efficiency and usability. To mitigate this manual parameter selection, herein, we propose AutoSOUL, a deep convolutional neural network (CNN)-based framework, to autonomously select the optimal regularization parameters of SOUL. A custom-designed CNN was trained on simulated datasets and fine-tuned on phantom datasets to classify a strain image as acceptable or unacceptable and predict the optimal strain image and parameter set corresponding to a given pre- and post-compression RF frame set. When tested on simulated, phantom, and <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">in vivo</i> breast datasets, the ranges of the accuracy, precision, recall, and F1 scores produced by the trained CNN model were 77-99%, 68-98%, 97-100%, and 81-99%, respectively. The RMSE, MAE, SNR, CNR, and SR values calculated from the optimal strain images suggested by AutoSOUL were within 1.5-4%, 6-9%, 0-30%, 0-21%, and 0-6%, respectively, of values obtained from corresponding manually tuned strain images. AutoSOUL reduced the execution time of the otherwise required manual tuning and selection process of SOUL by a factor of 84.4. These results indicate that AutoSOUL has the potential to enhance the usability of ultrasound elastography, facilitating more efficient clinical evaluation of tissue abnormalities.
Listening to the sound of light to guide surgeries and interventions
2026-04-01
article1st authorCorrespondingReal-time photoacoustic imaging with freehand light delivery and augmented reality guidance
Photoacoustics · 2026-04-01
articleOpen accessSenior author2026-03-05
article1st authorCorrespondingUltrasound in Medicine & Biology · 2026-01-23
articleOpen accessSenior authorBioengineering · 2026-03-29
articleOpen accessUltrasound imaging is a crucial tool for guiding radiation therapy, particularly for cancers such as pancreatic cancer, where tumors exhibit respiration-induced motion. While flexible ultrasound transducers offer improved anatomical conformity and reduced compression-induced distortion compared to rigid probes, their variable geometry presents significant challenges for conventional beamforming. In this study, we investigate a deep learning-based beamforming framework that directly predicts delayed RF data from raw RF input, bypassing explicit transducer shape estimation and traditional delay-and-sum computations. Building upon an artificial curvature simulation strategy, we systematically analyze the impact of curvature-induced variation and inherent RF noise on model performance and generalizability. We further introduce frequency-domain analysis to quantify RF-level signal variation that may not be apparent in spatial-domain image comparisons. Our results demonstrate that although noise-augmented training improves prediction consistency, reconstruction performance remains limited under the current prototype noise conditions. These findings highlight the importance of RF data diversity and noise characterization in developing clinically robust deep learning beamformers for flexible transducer-based ultrasound-guided radiation therapy.
2026-03-04
articleSenior authorPhotoacoustic imaging is promising for noninvasive breast cancer detection. However, previous simulation studies demonstrate diminished photoacoustic image quality with greater melanin content, depending on the use of delay-and-sum (DAS) or short-lag spatial coherence (SLSC) beamformers. Herein, we investigate the applicability of these findings to experimental data, including tissue-mimicking phantom data and in vivo breast data acquired from 20 patients with skin tone individual typology angles (ITAs) ranging from -78° (dark) to 71° (very light). Raw photoacoustic channel data were acquired with 757 nm and 1064 nm wavelengths and beamformed with DAS and SLSC. With brown and dark skin tones (i.e., ITA ≤ 0°) and targets ≥5 mm deep, the mean signal-to-noise ratios of simulated, phantom, and in vivo targets were 2.07, 12.67, and 19.41, respectively, with SLSC beamforming, relative to 1.97, 7.84, and 10.12 with DAS beamforming. Therefore, SLSC beamforming has the potential to advance equity across skin tones in photoacoustic breast imaging, relative to the skin tone bias observed with amplitude-based beamforming.
Tracking performance and surgeon assessment of photoacoustic-based surgical guidance methods
International Journal of Computer Assisted Radiology and Surgery · 2026-05-23
articleOpen accessSenior authorAbstract Purpose Photoacoustic imaging has the potential to provide real-time surgical guidance to avoid ureteral injury during hysterectomies, including photoacoustic-based distance measurements and auditory alerts to warn surgeons of unsafe proximity. Validation of tool tip tracking performance and assessments of surgeon interaction with multiple feedback options derived from photoacoustic-based tool tip tracking are essential to assess the expected impact. Methods Tracked tool tip positions were compared with ground-truth data acquired to validate our recently proposed velocity-based tracking approach. A user study with gynecological surgeons, OB/GYN residents, and a control group of non-medical professionals was then performed to determine the extent of improvements that can be achieved with multiple photoacoustic-based feedback techniques derived from the proposed velocity-based tracking approach. Results The average mean tracking errors across various programmed tool motions were 0.24 mm and 0.56 mm in the lateral and axial dimensions, respectively. Combinations of photoacoustic-based distance measurements, auditory feedback, and photoacoustic video display improved correct classifications of tool-to-ureter distances by 11.1% to 48.1% and reduced the aggregated response times among users by 1.72–2.15 s (i.e., 37–47% improvement) relative to camera videos. Conclusion Results are promising for the continued development of a photoacoustic surgical guidance system that successfully mitigates ureteral injuries during hysterectomies, simultaneously enabling reduced reliance on subjective camera-based visualization and addressing limitations of depth perception in conventional surgical imaging. This integration of photoacoustic image guidance into intraoperative workflows has the potential to enhance surgical safety.
Stroke · 2026-01-29
articleIntroduction: Shoulder pain occurs in 30-70% of patients with stroke and may persist in the chronic stage, interfering with motor recovery and quality of life. Reduced mobility and spasticity may result in the accumulation of hyaluronan (HA) in the extracellular matrix of muscles on the paretic side compared to non-paretic side, leading to dysfunctional gliding and reduced shoulder external rotation which is strongly correlated with post stroke shoulder pain (PSSP). Methods: 35 participants with PSSP and limitation in shoulder external rotation underwent MRI of the shoulder girdle musculature using T1, T2, T1ρ, and Dixon fat fraction imaging. HA concentration in the shoulder internal (pectoralis major) and external (infraspinatus) rotators was measured using T1ρ MRI relaxation time. Muscle atrophy and fat fraction were quantified using T1 closed surface volumetric analysis and Dixon imaging, respectively. Dynamic ultrasound shear strain mapping was used to quantify dysfunctional gliding of muscle. Pain was assessed using quantitative sensory testing. Results: T1ρ relaxation time was significantly increased in the infraspinatus on the paretic side (p) compared with the non-paretic (np) side (mean±SD: p, 29.1±3.0 vs. np, 27.2±2.2, Cohen’s d 0.75, p=0.002). Maximum shear strain was also reduced between the pectoralis major and minor muscles (p, 35.4±12.8% vs. np, 46.5±19.8%, Cohen’s d 0.66, p=0.008). Significant negative correlations were noted between maximum shear strain and T1ρ relaxation time difference between the paretic and non-paretic sides, (r=-0.6, p=0.0003), and pain severity (r=-0.5, p=0.03), suggesting that reduced muscle gliding is associated with increased HA accumulation and pain. There was greater atrophy in the infraspinatus (p, 94.1±45.6 vs. np, 122.1±54.1, p=0.0008), but fat fraction was not significantly increased. HA accumulation on T1ρ imaging was associated with greater muscle atrophy (r=-0.8, p=0.02). Conclusions: The results suggest that T1ρ MRI imaging is a valid structural biomarker for HA accumulation, and that ultrasound shear strain imaging is a valid functional biomarker for myofascial dysfunction. HA accumulation in paretic muscles may be a biologic cause of myofascial pain and dysfunction after stroke and is associated with muscle atrophy. Shoulder rehabilitation to prevent atrophy and HA accumulation in the shoulder muscles may disrupt the pathophysiology of post stroke shoulder pain.
Recent grants
CAREER: Technical & Theoretical Foundations for Photoacoustic-Guided Surgery
NSF · $540k · 2018–2025
NSF · $1.0M · 2024–2029
Minimizing Uncertainty in Breast Ultrasound Imaging with Real-Time Coherence-Based Beamforming
NIH · $1.6M · 2022–2027
Coherence-Based Photoacoustic Image Guidance of Transsphenoidal Surgeries
NIH · $726k · 2017–2021
NIH · $859k · 2020–2022
Frequent coauthors
- 225 shared
Patrick Y. Wen
- 175 shared
David A. Reardon
Dana-Farber Cancer Institute
- 175 shared
J. Raizer
Columbia University
- 125 shared
D. Schiff
University of Pittsburgh
- 125 shared
Thomas Kaley
Memorial Sloan Kettering Cancer Center
- 120 shared
Terri S. Armstrong
Patient-Centered Outcomes Research Institute
- 105 shared
Lisa M. DeAngelis
Memorial Sloan Kettering Cancer Center
- 100 shared
Timothy F. Cloughesy
University of California, Los Angeles
Awards & honors
- Trailblazer Award (2018) from the National Institute of Biom…
- MIT Technology Review, 35 Innovators Under 35 (2016)
- Alfred P. Sloan Research Fellow (2019)
- Maryland’s Outstanding Young Engineer (2019)
- Johns Hopkins Discovery Award (2018)
- Resume-aware match score
- Save to shortlist
- AI-drafted outreach
See your match with Muyinatu Bell
PhdFit ranks faculty by your research interests, methods, and publications — grounded in their actual work, not templates.
- Free to start
- No credit card
- 30-second signup