About

Yaochun Yu is an Assistant Professor of Civil and Environmental Engineering at Stanford University. His research focuses on functional environmental microbiology and environmental analytical chemistry, aiming to uncover and harness microorganisms for chemical biotransformation. He integrates high-resolution mass spectrometry, meta-omics sequencing, molecular microbiology and biochemistry, and computational modeling to identify the functional microbes, genes, and enzymes that drive these processes. Building on mechanistic insights, his work aims to develop environmentally benign chemicals and novel biosolutions for bioremediation and waste-to-resource recovery. He is also interested in how anthropogenic perturbations, such as chemical exposure, reshape microbial biodiversity and ecosystem function across natural and engineered ecosystems. His research seeks to resolve cause–effect relationships and, through the use of standardized and synthetic microbial communities, run hypothesis-driven experiments that translate fundamental insights into predictive tools and practical interventions. The overarching goal of his work is to help keep human activities within the safe operating space of planetary boundaries while advancing environmental and public health. Yaochun Yu earned his Ph.D. and M.S. in Environmental Science and Engineering from the University of Illinois Urbana-Champaign in 2021 and 2017, respectively. He completed his B.Eng. in Hydrology and Water Resources Engineering at Jilin University in 2015. He is based at Stanford University, where he is actively engaged in research and teaching within the Department of Civil and Environmental Engineering.

Research topics

• Cardiology
• Medicine
• Artificial Intelligence
• Computer Science
• Internal medicine
• Nuclear medicine
• Surgery
• Radiology

Selected publications

• Both Infarcted and Noninfarcted Brain Regions Contribute to Deep Learning–Based MRI Prediction of Acute Stroke Outcome

  American Journal of Neuroradiology · 2025-11-06

  articleOpen access

  <h3>BACKGROUND AND PURPOSE:</h3> Predicting long-term clinical outcomes based on early acute ischemic stroke (AIS) information would be useful for many reasons, including patient counseling and clinical trial execution. This study investigates how different regions in brain imaging, including noninfarcted areas, contribute to the accuracy of predicting 90-day stroke outcomes by using deep learning (DL). <h3>MATERIALS AND METHODS:</h3> We developed and validated DL models in 449 patients with AIS, by using MRI DWI scans from 1–7 days poststroke and 90-day mRS outcome data. These models were trained on various inputs: infarct volumes, full-brain images, infarct masks, intensity-preserved infarct masks, and images in which the infarct region is removed, which we call lesion-neutralized images. Performance was assessed by using accuracy of predicting the specific mRS score, accuracy within ±1 mRS category, mean absolute error (MAE), and the area under the curve (AUC) to predict unfavorable outcome (mRS &gt; 2). <h3>RESULTS:</h3> The model trained by using only infarct volume size reported the highest (worst) MAE of 1.51 points (95% CI, 1.40–1.61; <i>P</i> &lt; .001), while the model trained with full-brain images achieved the lowest MAE of 1.07 points (95% CI, 0.99–1.16). Models with intermediate amounts of imaging information each improved on the volume-only predictions but did not reach the performance of the full brain images; infarct masks, intensity-preserved infarct masks, and lesion-neutralized images demonstrated MAEs of 1.25 (95% CI, 1.15–1.34; <i>P</i> = .002), 1.21 (95% CI, 1.11–1.30; <i>P</i> = .008), and 1.35 (95% CI, 1.24–1.45; <i>P</i> &lt; .001), respectively. Similar results were seen for other prediction tasks, including AUC to predict unfavorable outcomes, ranging from 0.68 (95% CI, 0.63–0.73) for infarct volume to 0.86 (95% CI, 0.82–0.89) for full brain inputs. <h3>CONCLUSIONS:</h3> While the best performance came from by using the full brain imaging volume, we demonstrate that the infarct location, its signal characteristics, and importantly, the noninfarcted regions all contribute to the predictions. The noninfarcted areas may be a proxy for overall brain health and resilience, containing important information about potential outcomes.

  PublisherDOI

• Chronic cerebral hypoperfusion‐induced myelin loss in normal‐appearing white matter correlates with cognitive decline: insights from moyamoya disease

  Alzheimer s & Dementia Translational Research & Clinical Interventions · 2025-10-01 · 1 citations

  articleOpen access

  Abstract INTRODUCTION Chronic cerebral hypoperfusion (CCH)‐induced white matter hyperintensities (WMHs) are a well‐established risk factor for cognitive impairment and dementia. While animal and post mortem studies suggest that myelin loss in normal‐appearing white matter (NAWM) precedes WMHs, in vivo evidence in human brain remains limited. We aimed to investigate the associations between myelin changes in NAWM, CCH, and cognitive function in patients with moyamoya disease (MMD, a human model of CCH). METHODS We included 58 patients with MMD and 36 healthy controls, and all participants underwent 3.0T magnetic resonance imaging (MRI) with T1‐weighted, T2‐weighted, and arterial spin labeling (ASL) sequences. Myelin was assessed by the T1‐weighted/T2‐weighted ratio (rT1/T2), and cerebral blood flow (CBF) in each arterial region was measured by ASL. Cognitive function was evaluated using the Mini‐Mental State Examination (MMSE) and Montreal Cognitive Assessment (MoCA). RESULTS Patients exhibited decreased rT1/T2 at lower percentiles (P5 and P10) and increased rT1/T2 at higher percentiles (P75, P90, and P95) in NAWM (false discovery rate [FDR]‐corrected p &lt; 0.005), along with reduced CBF in the region of anterior circulation (FDR‐corrected p &lt; 0.05). Voxel‐based analysis of NAWM showed region‐specific decreases and increases in rT1/T2 in MMD. The percentile ranges (P90‐P10 and P95‐P5) of rT1/T2 showed high accuracy in differentiating patients from controls (accuracy: 88% to 90%). Multivariate analysis further revealed that in patients, both P5 and P10 of rT1/T2 were significantly associated with reduced CBF in the anterior circulation region (standardized β = 0.318, p = 0.016 and standardized β = 0.267, p = 0.043), and P5 of rT1/T2 significantly correlated with MMSE and MoCA (standardized β = 0.332, p = 0.004 and standardized β = 0.260, p = 0.011). DISCUSSION Our study provides in vivo evidence that CCH induces both myelin loss and potential plastic adaptations in NAWM, with severe myelin loss being associated with cognitive decline. Highlights MMD features CCH and cognitive decline. Decreased rT1/T2 (myelin loss) in NAWM was detected in MMD. The lower percentage of rT1/T2 correlated with CCH. The lowest percentage of rT1/T2 correlated with cognitive decline. The rT1/T2 showed high accuracy in differentiating patients from controls.

  PublisherOA PDFDOI

• Abstract 69: Both Infarct and Non-infarcted Brain Regions Drive Acute Deep Learning Based MRI Prediction of Stroke Outcome

  Stroke · 2025-01-30

  article

  Purpose: Predicting long-term outcomes from early acute ischemic stroke (AIS) information is crucial for prognostication, resource management, and clinical trials. Current methods predominantly depend on infarct-related features, such as size and location, for outcome predictions, and demonstrate limited correlation with outcomes. This study examines how deep learning leverages different brain regions, including non-infarct areas, to improve the accuracy of 90-day outcome predictions in AIS patients. Materials and Methods: We developed and validated deep learning (DL) models on a cohort of 436 AIS patients, using MRI diffusion-weighted imaging scans from 1-7 days post-stroke and 90-day modified Rankin Scale (mRS) follow-up data. These models were trained on various inputs— infarct volumes, whole-brain images, infarct masks, intensity-preserved infarct masks, and images in which the infarct region is removed, which we call infarct-neutralized images, and which enable an assessment of overall brain health. Model performance was assessed based on the accuracy of predicting the specific mRS score, accuracy within ±1 mRS category, mean absolute error (MAE), and the ability to predict unfavorable outcomes (mRS &gt; 2) using receiver operator curve (ROC) metrics. Results: The infarct volume model had the highest (worst) MAE of 1.48 points (95% CI: 1.38-1.58, p &lt; 0.001), while the whole-brain model achieved the lowest (best) MAE of 1.08 points (95% CI: 1.00-1.17). Models with intermediate imaging information—such as infarct masks (MAE 1.28, 95% CI: 1.19-1.37, p = .002), intensity-preserved infarct masks (MAE 1.27, 95% CI: 1.18-1.37, p = .009), and infarct-neutralized images (MAE 1.33, 95% CI: 1.24-1.43, p &lt; .001)—improved upon the volume-only predictions. For predicting unfavorable outcomes, the infarct volume model had the lowest performance (AUC 0.70, 95% CI: 0.65-0.75; p &lt; .001), while the whole-brain model achieved the highest AUC of 0.85 (95% CI: 0.82-0.89), outperforming the infarct mask (AUC 0.80, 95% CI: 0.76-0.84; p = .01), intensity-preserved infarct mask (AUC 0.80, 95% CI: 0.76-0.84; p = .007), and infarct-neutralized images (AUC 0.74, 95% CI: 0.69-0.79; p &lt; .001) Conclusions: The best predictive performance was achieved using voxel values from the entire brain, showing that both infarcted and non-infarcted regions contribute significantly to accuracy. Non-infarcted areas may reflect overall brain health and resilience, informing potential outcomes.

  PublisherDOI

• The Sensitivity of Arterial Spin-Labeling Imaging for Detection of Head and Neck Paragangliomas.

  PubMed · 2025-07-01

  articleOpen access1st authorCorresponding

  BACKGROUND AND PURPOSE: Head and neck paragangliomas (HNPGs) are rare neuroendocrine tumors whose hypervascular nature allows differentiation from many other head and neck neoplasms. We aimed to investigate the sensitivity of arterial spin-labeling (ASL) MR sequences for the detection of HNPGs. MATERIALS AND METHODS: All head and neck MR examinations performed at a single tertiary institution between 2015 and 2023 were searched. Studies using ASL sequences that indicated either clinical suspicion for or ultimate imaging diagnosis of HNPG were identified. These studies were independently reviewed by 2 neuroradiologists blinded to the original radiology reports to determine, in a stepwise fashion, the following: 1) whether there was asymmetrically elevated blood flow on ASL imaging, 2) whether ASL findings correlated with lesions identifiable on conventional anatomic images, and 3) whether lesions likely reflected paragangliomas on the basis of correlations with clinical, laboratory, pathology, and other radiology data (Disagreement between raters was resolved by consensus.). The Cohen κ coefficient and the sensitivity of ASL in identifying HNPGs were calculated. RESULTS: < .001). Among 18 cases with pathology- or dotatate PET-proved HNPG, the sensitivity was 100% for reader A and 94% for reader B. CONCLUSIONS: Asymmetrically elevated blood flow on ASL imaging demonstrates high sensitivity for the detection of HNPG, with almost perfect interrater agreement.

  PublisherOA PDFDOI

• Sex-Based Differences in Vasodilatory Flow Capacity Measured with 4D Flow MRI and White Matter Disease Burden Among PWH: A Preliminary Study

  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: Female individuals living with HIV (PWH) have a significantly higher risk of ischemic stroke than uninfected individuals, yet most studies on cerebrovascular health in PWH lack sex-specific focus. Differences in vasodilatory flow capacity (VFC) between sexes remain underexplored. Goal(s): To assess sex-stratified VFC differences and investigate its relationship with white matter disease burden in PWH. Approach: Using 4D Flow MRI, VFC was measured pre- and post-acetazolamide in 26 PWH patients, with correlation testing for white matter (WM) disease burden. Results: Higher median VFC was observed in female PWH, with regional variability, and an inverse association with WM disease burden in male PWH. Impact: Our findings underscore the need for sex-specific studies in cerebrovascular health for PWH, potentially guiding targeted interventions for neurovascular and cognitive health. Further research in larger cohorts could validate these preliminary observations and clarify their clinical implications.

  PublisherDOI

• Multi-scale wavelet decomposition and feature fusion for rotating machinery fault diagnosis under multi-level class imbalance

  Mechanical Systems and Signal Processing · 2025-09-30 · 8 citations

  article
  PublisherDOI

• Correlation between 4D-CT Enhancement Patterns and Parathyroid Adenoma Cellularity: A Retrospective Study

  American Journal of Neuroradiology · 2025-11-07 · 1 citations

  articleOpen access

  <h3>ABSTRACT</h3> <h3>BACKGROUND AND PURPOSE:</h3> Four-dimensional computed tomography (4DCT) is a well-established modality for preoperative localization of parathyroid adenomas^1-3. However, the relationship between enhancement patterns and histologic features such as lesion cellularity remains unclear. This study evaluates whether enhancement patterns correlate with lesion cellularity in surgically confirmed parathyroid adenomas. <h3>MATERIALS AND METHODS:</h3> A retrospective cohort study was performed on 134 patients (194 parathyroid lesions) who underwent 4DCT for preoperative localization and subsequent parathyroidectomy. After excluding lesions that were not visualized on 4DCT, not resected, or lacked pathology data on cellularity, 121 lesions in 95 patients were included in the final analysis. Enhancement patterns were categorized into Type A, B, or C based on previously published criteria⁴. Cellularity was dichotomized into increased and non-increased groups using an 80% cutoff. Mixed effect logistic regression analysis was performed to adjust for potential confounders. <h3>RESULTS:</h3> Among the 121 lesions with reported enhancement pattern and cellularity, 26 (21.5%) exhibited Type A enhancement, 65 (53.7%) Type B, and 30 (24.8%) Type C. Increased cellularity was observed in 88 lesions (72.7%). Type A enhancement was independently associated with increased cellularity (OR 6.1, 95% CI 1.3-29.3, p=0.02) after adjusting for confounders. Larger lesion size was also a significant predictor of increased cellularity (p&lt;0.001). No significant correlation was found between enhancement category B or C and cellularity. <h3>CONCLUSIONS:</h3> Type A enhancement on 4DCT is independently associated with increased cellularity in parathyroid adenomas. These findings suggest that enhancement kinetics may serve as a surrogate marker for histologic features, providing additional information for preoperative planning and surgical decision-making. ABBREVIATIONS: 4DCT = four-dimensional contrast-enhanced computed tomography neck; MGD = multi-gland disease; pHPT = primary hyperparathyroidism; PTH = parathyroid hormone; SPECT = single photon emission-computed tomography

  PublisherDOI

• Exploring the Impact of Hemoglobin on Cerebral Blood Flow in Arterial Territories and Surgical Outcomes: Potential Implications for Moyamoya Disease Treatment

  Journal of the American Heart Association · 2024-09-30 · 4 citations

  articleOpen access

  Background Changes in levels of hemoglobin would result in alterations of cerebral blood flow (CBF). However, the impact of hemoglobin on CBF in moyamoya disease (MMD) remains largely unknown. This study sought to determine whether CBF would be influenced by hemoglobin before surgical revascularization and to analyze the relationships between hemoglobin and CBF with clinical outcome after surgery in patients with MMD. Methods and Results We prospectively enrolled adult patients with MMD undergoing surgical revascularization between June 2020 and December 2022. Preoperative CBF was measured in the territories of anterior, middle, and posterior cerebral arteries (ACA, MCA, and PCA, respectively) using 3‐dimensional pseudo‐continuous arterial spin labeling magnetic resonance imaging. Clinical outcome at 1 year after surgery was evaluated using the modified Rankin Scale. A total of 60 patients with MMD were included, with 25% (n=15) experiencing unfavorable outcomes. Patients with MMD exhibited lower CBF (ACA: P =0.007; MCA: P &lt;0.001; PCA: P =0.014), compared with healthy controls (n=40). Hemoglobin was negatively and significantly associated with CBF (ACA: β=−0.45, P &lt;0.001; MCA: β=−0.38, P &lt;0.001; PCA: β=−0.54, P &lt;0.001). CBF rather than hemoglobin was significantly related with clinical outcome (ACA: P &lt;0.001; MCA: P &lt;0.001; PCA: P =0.001), and CBF showed high discrimination in predicting clinical outcome (ACA: area under the curve, 0.84; MCA: area under the curve, 0.84; PCA: area under the curve, 0.80). Conclusions Our findings demonstrate that hemoglobin significantly influences CBF, and CBF has a high predictive value for clinical outcome in MMD. The optimal hemoglobin level before surgical revascularization should be further investigated.

  PublisherDOI

• A Clinical and Imaging Fused Deep Learning Model Matches Expert Clinician Prediction of 90-Day Stroke Outcomes

  American Journal of Neuroradiology · 2024-02-08 · 13 citations

  articleOpen access

  <h3>BACKGROUND AND PURPOSE:</h3> Predicting long-term clinical outcome in acute ischemic stroke is beneficial for prognosis, clinical trial design, resource management, and patient expectations. This study used a deep learning–based predictive model (DLPD) to predict 90-day mRS outcomes and compared its predictions with those made by physicians. <h3>MATERIALS AND METHODS:</h3> A previously developed DLPD that incorporated DWI and clinical data from the acute period was used to predict 90-day mRS outcomes in 80 consecutive patients with acute ischemic stroke from a single-center registry. We assessed the predictions of the model alongside those of 5 physicians (2 stroke neurologists and 3 neuroradiologists provided with the same imaging and clinical information). The primary analysis was the agreement between the ordinal mRS predictions of the model or physician and the ground truth using the Gwet Agreement Coefficient. We also evaluated the ability to identify unfavorable outcomes (mRS &gt;2) using the area under the curve, sensitivity, and specificity. Noninferiority analyses were undertaken using limits of 0.1 for the Gwet Agreement Coefficient and 0.05 for the area under the curve analysis. The accuracy of prediction was also assessed using the mean absolute error for prediction, percentage of predictions ±1 categories away from the ground truth (±1 accuracy [ACC]), and percentage of exact predictions (ACC). <h3>RESULTS:</h3> To predict the specific mRS score, the DLPD yielded a Gwet Agreement Coefficient score of 0.79 (95% CI, 0.71–0.86), surpassing the physicians’ score of 0.76 (95% CI, 0.67–0.84), and was noninferior to the readers (<i>P </i>&lt; .001). For identifying unfavorable outcome, the model achieved an area under the curve of 0.81 (95% CI, 0.72–0.89), again noninferior to the readers’ area under the curve of 0.79 (95% CI, 0.69–0.87) (<i>P </i>&lt; .005). The mean absolute error, ±1ACC, and ACC were 0.89, 81%, and 36% for the DLPD. <h3>CONCLUSIONS:</h3> A deep learning method using acute clinical and imaging data for long-term functional outcome prediction in patients with acute ischemic stroke, the DLPD, was noninferior to that of clinical readers.

  PublisherOA PDFDOI

• Random expert sampling for deep learning segmentation of acute ischemic stroke on non-contrast CT

  Journal of NeuroInterventional Surgery · 2024-02-01 · 10 citations

  articleOpen access

  Background Outlining acutely infarcted tissue on non-contrast CT is a challenging task for which human inter-reader agreement is limited. We explored two different methods for training a supervised deep learning algorithm: one that used a segmentation defined by majority vote among experts and another that trained randomly on separate individual expert segmentations. Methods The data set consisted of 260 non-contrast CT studies in 233 patients with acute ischemic stroke recruited from the multicenter DEFUSE 3 (Endovascular Therapy Following Imaging Evaluation for Ischemic Stroke 3) trial. Additional external validation was performed using 33 patients with matched stroke onset times from the University Hospital Lausanne. A benchmark U-Net was trained on the reference annotations of three experienced neuroradiologists to segment ischemic brain tissue using majority vote and random expert sampling training schemes. The median of volume, overlap, and distance segmentation metrics were determined for agreement in lesion segmentations between (1) three experts, (2) the majority model and each expert, and (3) the random model and each expert. The two sided Wilcoxon signed rank test was used to compare performances (1) to 2) and (1) to (3). We further compared volumes with the 24 hour follow-up diffusion weighted imaging (DWI, final infarct core) and correlations with clinical outcome (modified Rankin Scale (mRS) at 90 days) with the Spearman method. Results The random model outperformed the inter-expert agreement ((1) to (2)) and the majority model ((1) to (3)) (dice 0.51±0.04 vs 0.36±0.05 (P &lt; 0.0001) vs 0.45±0.05 (P &lt; 0.0001)). The random model predicted volume correlated with clinical outcome (0.19, P &lt; 0.05), whereas the median expert volume and majority model volume did not. There was no significant difference when comparing the volume correlations between random model, median expert volume, and majority model to 24 hour follow-up DWI volume (P &gt; 0.05, n=51). Conclusion The random model for ischemic injury delineation on non-contrast CT surpassed the inter-expert agreement ((1) to (2)) and the performance of the majority model ((1) to (3)). We showed that the random model volumetric measures of the model were consistent with 24 hour follow-up DWI.

  PublisherOA PDFDOI

Frequent coauthors

• David S. Liebeskind

  44 shared

• Greg Zaharchuk

  43 shared

• Gregory W. Albers

  Stanford Medicine

  41 shared

• Maarten G. Lansberg

  Palo Alto University

  34 shared

• Weihai Xu

  South China Sea Institute Of Oceanology

  27 shared

• Jiahong Ouyang

  Stanford University

  27 shared

• Mingli Li

  Peking Union Medical College Hospital

  26 shared

• Yongkai Liu

  Nanjing University of Aeronautics and Astronautics

  19 shared

Education

• Ph.D., Civil and Environmental Engineering

  Stanford University

  2017

• M.S., Civil and Environmental Engineering

  University of California, Berkeley

  2012

• B.S., Environmental Engineering

  Tsinghua University

  2009