Upload your resume. PhdFit's six research agents compare your background with faculty profiles, recent publications, lab focus, and outreach opportunities, then rank professors with evidence you can review.
Resume-aware matchingEvidence from recent researchOutreach drafts included
Advisor Match Brief Resume uploaded
Resume signals5 found
NLPHCIFairnessGraph MLPyTorch
Six-agent analysisRunning
Axel · Candidate AnalystResume parsed
Nova · Professor ResearcherRecent papers
Echo · Match ExplainerEvidence linked
Atlas · Strategy AdvisorFit ranked
Lyra · Outreach WriterDraft queued
Juno · Meeting Prep CompilerNext step ready
SC
Dr. Sarah Chen
Stanford · NLP & interpretability
91
fit
Fairness + NLP projects overlap with her recent work.
Strong methods fit: model evaluation and interpretability.
Clear outreach angle for current trustworthy AI research.
Outreach angle
Ask how her lab is extending interpretability methods into fairness audits for real-world AI systems.
Jeremy Dahl is an Associate Professor of Radiology (Pediatric Radiology) at Stanford University. He is affiliated with the Center for Artificial Intelligence in Medicine & Imaging (AIMI). His role involves advancing research in artificial intelligence applications within the field of pediatric radiology, contributing to the development of innovative imaging techniques and AI-driven healthcare solutions. His work is focused on integrating AI technologies into medical imaging to improve diagnostic accuracy and patient outcomes, leveraging his expertise in radiology and AI to foster interdisciplinary collaboration and innovation in medical imaging.
Pulse-echo speed-of-sound imaging is emerging as a powerful extension of conventional echography. It shows particular promise for diagnosing metabolic dysfunction-associated steatotic liver disease, whose prevalence is increasing worldwide alongside obesity rates. However, current techniques often lack robustness, particularly when limited data are available. Existing methods rely on different regularization strategies to incorporate knowledge on the speed-of-sound distribution within tissues. In this work, we propose a novel regularization approach that automatically integrates information from the B-mode image. The proposed regularizer relies on the principle of sparsity with partially known support applied to a total variation norm, enabling increased robustness compared to regularizers disregarding information contained in B-mode images, or circumventing drawbacks of approaches based on the manual segmentation of the latter. We evaluate the proposed technique with simulated and in vivo data, considering both large and reduced datasets, with a specific focus on liver imaging. The results indicate that our method enhances speed-of-sound accuracy with simulated data when compared with regularizers proposed in the literature, whereas the robustness to data reduction is improved with in vivo data. Notably, our approach can facilitate the deployment of pulse-echo speed-of-sound imaging on ultra-portable transducers, which are typically constrained in data acquisition capacity.
In pulse-echo ultrasound, aberration often degrades image quality when beamforming does not account for wavefront distortions. To address this issue, local sound speed estimators have been developed in the past decade for distributed aberration correction. Recently, methods based on iterative optimization have improved sound speed accuracy with respect to earlier approaches. However, the accuracy of these newer methods is limited by media with reverberation clutter and by the straight-ray model of wave propagation. To address these challenges, we propose using wavefield correlation (WFC) beamforming when performing sound speed optimization. WFC, an ultrasound adaptation of reverse time migration, correlates simulated forward-propagated transmit wavefields and backwards-propagated receive wavefields in order to reconstruct images. This process more accurately models wave propagation in heterogeneous media and can decrease diffuse clutter due to its spatiotemporal matched filtering effect. We implement herein a WFC beamformer using an auto-differentiation software and estimate the sound speed map by optimizing a regularized common-midpoint phase focusing criterion using gradient descent. This approach is compared to a previous method relying on delay and sum (DAS) with straight-ray time delay calculations on a variety of simulated, phantom, and in vivo data with large sound speed variations and clutter. Results show that using WFC decreases sound speed estimation error, leading to improvements in resolution and contrast in the corrected image. In particular, these promising results have potential to improve pulse-echo imaging for challenging clinical scenarios.
In pulse-echo ultrasound, aberration often degrades image quality when beamforming does not account for wavefront distortions. To address this issue, local sound speed estimators have been developed in the past decade for distributed aberration correction. Recently, methods based on iterative optimization have improved sound speed accuracy with respect to earlier approaches. However, the accuracy of these newer methods is limited by media with reverberation clutter and by the straight-ray model of wave propagation. To address these challenges, we propose using wavefield correlation (WFC) beamforming when performing sound speed optimization. WFC, an ultrasound adaptation of reverse time migration, correlates simulated forward-propagated transmit wavefields and backwards-propagated receive wavefields in order to reconstruct images. This process more accurately models wave propagation in heterogeneous media and can decrease diffuse clutter due to its spatiotemporal matched filtering effect. We implement herein a WFC beamformer using an auto-differentiation software and estimate the sound speed map by optimizing a regularized common-midpoint phase focusing criterion using gradient descent. This approach is compared to a previous method relying on delay and sum (DAS) with straight-ray time delay calculations on a variety of simulated, phantom, and in vivo data with large sound speed variations and clutter. Results show that using WFC decreases sound speed estimation error, leading to improvements in resolution and contrast in the corrected image. In particular, these promising results have potential to improve pulse-echo imaging for challenging clinical scenarios.
IEEE Transactions on Ultrasonics · 2025-12-05 · 1 citations
article
Transabdominal ultrasound imaging requires acoustic wave propagation through the abdominal wall, which consists of a complex organization of tissue layers that degrade imaging of underlying organs such as the liver. A <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">priori</i> knowledge of the acoustic properties of this medium could enable adaptive imaging techniques, but the underlying ground truth properties in a living patient is a fundamentally unknown quantity. Numerical simulations can offer a solution to this challenge by modeling the propagation of ultrasound imaging pulses in media with known acoustic properties and can therefore provide a quantitative relationship between the received ultrasound echoes and the underlying tissue structure. However, to be comparable with clinical scenarios, <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">in silico</i> phantoms of the abdominal wall must be geometrically and quantitatively accurate, with rigorous tolerances. Here, we describe a dataset generation approach to produce experimentally calibrated ultrasound data for transabdominal imaging. Semantically labeled 3D <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">in silico</i> phantoms with 0.33mm isotropic resolution, along with physical parameters from the literature of the human abdominal wall, are used to model acoustic propagation through human tissue. We perform 2D numerical simulations of the Westervelt equation to generate a dataset of raw ultrasound data for liver imaging consisting of 1027 samples. We then discuss the applications of this dataset for facilitating abdominal ultrasound imaging research. The labeled abdominal dataset, simulated data, and the simulation tools that model wave propagation are made publicly available.
Ultrasound molecular imaging (UMI) is an advanced imaging modality that shows promise in detecting cancer at early stages. It uses microbubbles as contrast agents, which are functionalized to bind to cancer biomarkers overexpressed on endothelial cells. A major challenge in UMI is isolating bound microbubble signal, which represents the molecular imaging signal, from that of free-floating microbubbles, which is considered background noise. In this work, we propose a fast GPU-based method using robust principal component analysis (RPCA) to distinguish bound microbubbles from free-floating ones. We explore the method using simulations and measure the accuracy using the Dice coefficient and RMS error as functions of the number of frames used in RPCA reconstruction. Experiments using stationary and flowing microbubbles in tissue-mimicking phantoms were used to validate the method. Additionally, the method was applied to data from ten transgenic mouse models of breast cancer development, injected with B7-H3-targeted microbubbles, and two mice injected with non-targeted microbubbles. The results showed that RPCA using 20 frames achieved a Dice score of 0.95 and a computation time of 0.2 seconds, indicating that 20 frames is potentially suitable for real-time implementation. On in vivo data, RPCA using 20 frames achieved a Dice score of 0.82 with DTE, indicating good agreement between the two, given the limitations of each method.
Ultrasound molecular imaging (UMI) offers enhanced accuracy in early cancer detection, surpassing the capabilities of conventional ultrasound. UMI uses targeted microbubbles to detect specific molecular biomarkers expressed on vascular endothelial cells, where accurate assessment of molecular expression relies on capturing signals only from the targeted microbubbles that are bound to the specific biomarker. Elevated vascular density, however, can potentially amplify the molecular signal due to the increased availability of biomarker, thereby confounding the molecular signal readout. Elevated vascular density can also lead to an increased number of freely circulating microbubbles. In this work, we present a novel UMI metric based on the separation of signals from bound and free microbubbles to normalize the expression level of molecular signal based on vascular density. We base our metric on Robust Principal Component Analysis (RPCA), which we have previously used to separate bound and unbound microbubbles. Using data from a transgenic mouse model of spontaneous breast cancer, we show that the proposed metric quantifies targeted microbubble binding and remains between 2 and 3 across different mice within the same model regardless of vascular density. Furthermore, this ratio effectively distinguishes between cases with targeted and non-targeted microbubbles.
Ultrasound molecular imaging (USMI) is an approach that detects targeted microbubbles (MBs) that are bound to specific biomarkers of disease; however, effectively suppressing free-floating MBs remains challenging. In this study, we adopted 3D transformer to analyze time-series ultrasound dataset to achieve enhanced differentiation of bound MBs from free-floating counterparts.We present MBFormer, or microbubble transformer, a nondestructive USMI transformer model that processes time-series 3D spatio-temporal ultrasound videos. The architecture of our model consists of a hierarchical transformer framework that comprises (1) an encoder with the attention mechanism but without positional embeddings, (2) a lightweight decoder featuring multilayer perceptrons (MLPs), and (3) a probability head. The model was validated through an in vivo study employing a transgenic mouse model of spontaneous breast cancer. B-mode and contrast enhanced ultrasound (CEUS) videos were captured utilizing the Vevo2100 system (FUJIFILM Visualsonics, Toronto, Canada). The ground truth data was generated via robust principal component analysis (RPCA)-processed differential targeted enhancement (DTE).We assessed the performance of MBFormer against the baseline models: a CNN-based USMI approach and SegFormer3D, an established 3D transformer model for medical image segmentation. MBFormer demonstrated improved detectability of bound MBs, achieving an area under the curve (AUC) of 0.951, a correlation coefficient (CC) of 0.981, and a continuous dice coefficient (CDC) of 0.748, surpassing the baseline models. These findings suggest that our 3D time-series data processing using the proposed transformer architecture effectively suppresses free-floating MBs while enhancing the detectability of bound MBs.
Annual Review of Biomedical Engineering · 2025-02-19 · 3 citations
reviewOpen access
Medical ultrasound is a diagnostic imaging modality used for visualizing internal organs; the frequencies typically used are 2-10 MHz. Scanning acoustic microscopy (SAM) is a form of ultrasound where frequencies typically exceed 50 MHz. Increasing the acoustic frequency increases the specimen's spatial resolution but reduces the imaging depth. The advantages of using SAM over conventional light and electron microscopy include the ability to image cells and tissues without any preparation that could kill or alter them, providing a more accurate representation of the specimen. After scanning the specimen, acoustic signals are merged into an image on the basis of changes in the impedance mismatch between the immersion fluid and the specimens. The acoustic parameters determining the image quality are absorption and scattering. Surface scans can assess surface characteristics of the specimen. SAM is also capable of elastography, that is, studying elastic properties to discern differences between healthy and affected tissues. SAM has significant potential for detection/analysis in research and clinical studies.
IEEE Transactions on Medical Imaging · 2025-09-09 · 3 citations
articleOpen access
In ultrasound imaging, propagation of an acoustic wavefront through heterogeneous media causes phase aberrations that degrade the coherence of the reflected wavefront, leading to reduced image resolution and contrast. Adaptive imaging techniques attempt to correct this phase aberration and restore coherence, leading to improved focusing of the image. We propose an autofocusing paradigm for aberration correction in ultrasound imaging by fitting an acoustic velocity field to pressure measurements, via optimization of the common midpoint phase error (CMPE), using a straight-ray wave propagation model for beamforming in diffusely scattering media. We show that CMPE induced by heterogeneous acoustic velocity is a robust measure of phase aberration that can be used for acoustic autofocusing. CMPE is optimized iteratively using a differentiable beamforming approach to simultaneously improve the image focus while estimating the acoustic velocity field of the interrogated medium. The approach relies solely on wavefield measurements using a straight-ray integral solution of the two-way time-of-flight without explicit numerical time-stepping models of wave propagation. We demonstrate method performance through in silico simulations, in vitro phantom measurements, and in vivo mammalian models, showing practical applications in distributed aberration quantification, correction, and velocity estimation for medical ultrasound autofocusing.
