About

Kunal Agrawal is an Associate Clinical Professor in the Neurosciences School at UC San Diego. His research focuses on stroke care, neurovascular diseases, and the implementation of innovative technologies to improve stroke outcomes. He has contributed to various clinical trials and publications related to stroke diagnosis, treatment, and telemedicine, including the use of hologram teleportation for stroke assessment and transcranial Doppler in predicting post-treatment outcomes. His work also encompasses the development of stroke care models during the COVID-19 pandemic, the evaluation of diagnostic imaging techniques, and the application of augmented intelligence in stroke management. Agrawal's research includes systematic reviews and meta-analyses on cerebrovascular events, patent foramen ovale closure, and the cost-effectiveness of stroke workup procedures. His publications demonstrate a strong focus on advancing stroke diagnostics, treatment strategies, and healthcare delivery through technological integration and clinical research.

Research topics

• Internal medicine
• Medicine
• Cardiology
• Intensive care medicine
• Emergency medicine
• Surgery
• Physical therapy
• Physical medicine and rehabilitation
• Medical emergency

Selected publications

• Dietary omega-3 polyunsaturated fatty acids as a protective factor of myopia: the Hong Kong Children Eye Study

  British Journal of Ophthalmology · 2025-08-19 · 4 citations

  articleOpen access

  PURPOSE: To evaluate the associations between omega-3 polyunsaturated fatty acids (ω-3 PUFAs) and other dietary factors with myopia. METHODS: A total of 1005 Chinese children, aged from 6 to 8 years, from a population-based Hong Kong Children Eye Study, were included in the analysis. Diet was assessed using a validated food-frequency questionnaire. Cycloplegic spherical equivalent (SE) refraction was assessed with an autorefractometer, and axial length (AL) by an IOL Master. RESULTS: AL was longest in the lowest quartile group of ω-3 PUFAs intake, compared with the highest (adjusted mean (95% CI), 23.29 (23.17 to 23.40) mm vs 23.08 (22.96 to 23.19) mm, p=0.01; p-trend=0.02) after adjusting for age, sex, body mass index, near-work time, outdoor time, and parental myopia history. The corresponding trends were observed in SE (-0.13 (-0.32 to 0.07) D in the lowest and 0.23 (0.03 to 0.42) D in the highest quartile groups, p=0.01; p-trend=0.01). In contrast, AL was longest in the highest quartile group of saturated fatty acids (SFA) intake, compared with the lowest (23.30 (23.17 to 23.42) mm vs 23.13 (23.01 to 23.24) mm, p=0.05; p-trend=0.04). The corresponding trends were observed in SE (-0.12 (-0.33 to 0.09) D in the highest and 0.13 (-0.04 to 0.31) D in the lowest quartile group, p=0.06; p-trend=0.04). A lower intake of ω-3 PUFAs was associated with myopia (p-trend=0.006). None of the other nutrients were associated with SE or AL or myopia. CONCLUSIONS: Intake of ω-3 PUFAs is a protective factor against myopia, while higher SFA intake is a risk factor. Our findings indicated a possible effect of diet on myopia, of which ω-3 PUFAs intake may play a protective role against myopia development in children.

  PublisherOA PDFDOI

• Abstract TP78: TIEMPPO: Telestroke Influence of Educating Minority Patients and Providers to Optimize treatment times and stroke awareness

  Stroke · 2025-01-30

  article

  Introduction: Stroke education is critical for patients and healthcare providers managing acute ischemic stroke (AIS). In our hyperacute telestroke network, most spoke centers have shown improved key performance indicators (KPIs) for IV thrombolytics. However, some spoke centers with predominantly Spanish-speaking population and staff continue to face operational hurdles hindering KPI optimization. Previous bilingual education efforts in English and Spanish have shown limited success. We hypothesize that the cultural background of staff and patients at these sites may impact these efforts. Our goal was to explore the cultural factors influencing telestroke care to design an educational initiative that addresses the cultural context of our local community, with a focus on both cultural humility and competency, to improve AIS KPIs in telestroke networks. Methods: We conducted a descriptive, exploratory survey to identify key factors contributing to AIS KPIs, with a focus on cultural and language challenges. First, a survey was distributed online via RedCap to Emergency Department (ED) physicians, nurses, technicians, patients, and families. Survey data was analyzed to identify key themes. Based on these findings and future focus groups, culturally tailored educational lectures will be developed and presented. Time-sensitive KPIs will be tracked pre- and post-intervention using data from the Hub center’s clinical database. Results: From January 1st to September 12th, 2024, these spoke sites recorded overall KPI times slightly longer than the network median: Door-to-CT 21 minutes, Door-to-CT read 40 minutes, Door-to-stroke-code page activation 26 minutes, Door-to-decision 50 minutes, and Door-to-thrombolytic 68 minutes. To date, 25 ED staff members have completed the survey. Key findings include: 68% of staff are involved in Telestroke cases at least weekly, 80% feel prepared for Telestroke cases, and 68% expressed a need for post-stroke debriefings. Barriers to care included CT delays (44%) and difficulties with the Telestroke machine (44%). Additionally, 64% of staff indicated an immediate need for translation support. Further qualitative themes and updated KPI metrics will be reported. Conclusion: In a culturally diverse country like the U.S., medical conditions must consider the cultural backgrounds of both patients and care providers. Integrating cultural humility and competency into stroke care can enhance understanding and improve stroke outcomes.

  PublisherDOI

• Risk Prediction of Cerebrovascular Ischemic Events Following Cervical Artery Dissections Using High‐Intensity Transient Signals: A Systematic Review, Meta‐Analysis and a Single Center Experience

  Stroke Vascular and Interventional Neurology · 2025-03-08

  articleOpen access

  Background Predicting and managing spontaneous cervical artery dissections (CeAD) is challenging due to the absence of tools for early identification of high‐risk individuals. This study seeks to gather evidence on the predictive value of high‐intensity transient signals (HITS) detected by transcranial Doppler for recurrent ischemic events (IEs) following CeAD. Methods We performed a systematic review and meta‐analysis of published studies along with the data from our cohort. Following Preferred Reporting Items for Systematic Reviews and Meta‐Analyses guidelines, we searched PubMed, Embase, and Scopus to identify studies that evaluated HITS in patients with CeAD with the aim of predicting IEs. Data were pooled using a random effects model, with odds ratio (OR) and its 95% CI as the effect size. Heterogeneity was assessed with the Q statistic and I 2 test, and subgroup analysis evaluated the impact of dissected artery (carotid versus vertebral) on the relationship between HITS and IEs. Our retrospective study included consecutive patients diagnosed with CeAD, followed for 90 days to record IEs. Univariable and multivariable analyses were performed to identify factors associated with recurrent transient ischemic attacks or strokes within 90 days post CeAD. Results Our systematic review included 5 prior studies, which, combined with our center's sample size, provided data for a total of 306 patients. The meta‐analysis indicated that HITS is a significant predictor of IEs (OR: 13.25 [95% CI, 2.97–59.13], P <0.01) with low heterogeneity (I 2 = 42%, P = 0.13). However, subgroup analysis indicated that HITS are a significant predictor only for carotid artery dissections ( P <0.01) and not for vertebral artery dissections ( P = 0.11). The cohort consisted of 34 patients (mean age: 46.8 years, 55.9% male). The incidence of IEs was 20% in our cohort and all of them (100%) had HITSs in transcranial Doppler. In multivariable analysis, the presence of HITS ( P = 0.006) and intraluminal thrombosis ( P = 0.02) were significant predictors of IEs. Conclusion The presence of HITS detected by transcranial Doppler is a strong predictor of IEs in patients with carotid artery dissections. This highlights the clinical value of transcranial Doppler in identifying high‐risk patients and emphasizes the need for proactive management strategies to reduce the risk of future IEs in this subgroup.

  PublisherDOI

• PFO‐ACCESS: Augmenting Communications for Medical Care or Closure in the Evaluation of Patients With Stroke With Cardiac Shunts

  Stroke Vascular and Interventional Neurology · 2025-02-05

  articleOpen access

  Background Patent foramen ovale (PFO) contributes to a quarter of embolic strokes of undetermined source. Although the benefit of PFO closure in selected patients has been demonstrated, our system workflow still resulted in a low rate of PFO evaluation for closure. The aim of the PFO‐ACCESS (Augmenting Communications for Medical Care or Closure in the Evaluation of Stroke Patients With Cardiac Shunts) program (which included implementation of the Viz.ai PFO‐specific communications module) was to determine if there was any change in PFO management due to improved communication between stroke and interventional cardiology teams. Methods In this quality improvement project, we compared pre‐PFO ACCESS (December 2022–November 2023) to post‐PFO ACCESS periods (November 2023–June 2024) regarding PFO evaluations. The Viz.ai PFO module was implemented for the stroke and interventional cardiology teams without other workflow changes. Key performance indicators included referral frequency, PFO closure rates, and referral time intervals. Statistical comparisons utilized Mann–Whitney U , chi‐square, Fisher's exact, and exact Poisson test where appropriate. Results The postimplementation period noted a 492% PFO referral increase (11 versus 38,65 [annualized]; P <0.0001). PFO closure number totals showed a 186% nonsignificant increase pre versus post (6 versus 10,17 [annualized]; P = 0.99), with PFO closure of percentage of total referred cases showing a large but nonsignificant decrease (54.55%, 26.32%; P = 0.14). Time comparisons showed a marked but nonsignificant decrease in median “referral sent to referral viewed” (10:37 hours, 1:08 hours; P = 0.73), “referral sent to referral accepted” (10:37 hours, 1:03 hours; P = 0.67) time interval, and “referral sent to closure” time interval (102 days, 97 days; P = 0.55). Conclusion The PFO‐ACCESS program with Viz.ai PFO module use resulted in a 492% increase in PFO referrals due to enhanced communication and efficiency in managing PFO‐related stroke cases. Though the increased number of referrals and closures were observed, the PFO closure percentage of total referred cases showed a marked but nonsignificant decrease indicating selective case management. The higher number of PFO closures shows that more patients are indeed appropriate for PFO closure consideration. Future efforts should focus on expanding outpatient use and increasing provider education to optimize PFO management.

  PublisherDOI

• Abstract TP246: Transcranial Doppler (TCD) Parameters in Predicting Outcomes Following Successful Mechanical Thrombectomy of Large Vessel Occlusions in Anterior Circulation: A Systematic Review and Meta-Analysis

  Stroke · 2025-01-30

  review

  Introduction: Mechanical thrombectomy (MT) is a primary treatment for acute ischemic stroke due to large vessel occlusions. While effective, 20-40% of patients experience hemorrhagic transformation (HT), and around 50% fail to achieve favorable functional outcomes. Transcranial Doppler (TCD) is a non-invasive and cost-effective method for real-time monitoring of hemodynamic status following MT. However, the prognostic value of TCD parameters in predicting HT and poor functional outcome is unclear. We performed a systematic review and meta-analysis of 4 TCD parameters (mean flow velocity (MFV), MFV index, peak systolic velocity (PSV), and pulsatility index (PI) in patients with and without HT and favorable vs poor functional recovery (modified Rankin Scale (mRS) 0-2 vs 3-6). Methods: PubMed, Embase, and Scopus were searched on July 25, 2024 to identify observational studies in which TCD parameters were measured within 48 hours from successful MT (Thrombolysis in Cerebral Infarction 2b–3) of anterior circulation. Risk of bias assessment was performed using a standardized tool tailored for TCD studies. The standardized mean difference (Hedges’ g) with 95% CI and risk ratios (RRs) with 95% CI were calculated using random-effects models. The review was prospectively registered on PROSPERO (registration number CRD42024575381). Results: Eleven studies met inclusion criteria. No study had high risk of bias. MFV and MFV index were higher in patients with HT+ compared with HT- (Hedges' g = 0.42 and 0.54, p = 0.015 and 0.005, respectively). Patients with MFV index ≥1.3 showed a higher risk of all HT (RR = 2.01, 95% CI = 1.27–3.17, p = 0.003), symptomatic HT (RR, 4.68; 95% CI,1.49–14.65, p=0.008), and poor functional recovery at 90 days (RR, 1.66; 95% CI,1.32–2.08, p<0.001), respectively. There was no difference in mean PSV (p=0.1) and PI (p=0.3) among groups with and without HT. Conclusion: Our study highlights the prognostic value of TCD parameters, particularly MFV index, in predicting HT, symptomatic HT, and poor functional recovery after successful MT in the anterior circulation. Our findings were limited by low number of studies. Large-scale and multi-center studies are needed to confirm these findings and validate the MFV index as a reliable predictor to improve post-thrombectomy care.

  PublisherDOI

• Abstract TP333: Racial disparities in the management of newly diagnosed diabetes mellitus in Acute Ischemic Stroke

  Stroke · 2025-01-30

  articleSenior author

  Introduction: Diabetes Mellitus (DM) is a significant risk factor for acute ischemic stroke (AIS) and the incidence of newly diagnosed DM in AIS is higher in certain ethnicities. We aim to examine prescribing patterns of anti-hyperglycemic medications (anti-DM) in hospitalized AIS patients with newly diagnosed with DM. We also examine ethnic disparities in prescribing patterns amongst Hispanic and Non-Hispanic patients. Methods: We retrospectively examined prospectively collected data from an IRB-approved stroke registry at two academic Comprehensive Stroke Centers (CSCs). We included patients with a new diagnosis of AIS with no documented history of DM between 1/1/2013 and 6/30/2024. We examined baseline demographics, comorbidities, ethnicity, insurance status, acute stroke treatment, NIHSS, A1c, blood glucose and their association with anti-DM prescribing patterns. We also evaluated prescribing differences in Hispanic versus Non-Hispanic groups. Data was analyzed using correlation matrix, Pearson’s and Spearman’s correlation coefficients, and Chi-squared and t test, as appropriate. Results: A total of 2870 AIS patients were identified in the study period. Of these, 20.1% (n=578) were Hispanic, 42.5% (n=1219) were female, and 47.8% (n=1373) had Medicare as their payment source. New diagnosis of DM occurred in 6.0% (n=52/863), and 52.0% (n=27) were prescribed anti-DM at discharge. Overall, only a history of previous stroke was independently associated with prescribing anti-DM on discharge (p=0.03). Of patients with new DM, Hgb A1c (p=0.01) and blood glucose (p=0.006) were significantly associated with prescribing anti-DM on discharge. Patients with Medicaid (p=0.04) and no previous medical history (p=0.02) were less likely to receive anti-DM. Mean Hgb A1c was higher in patients that were prescribed anti-DM vs not prescribed anti-DM (8.82 vs. 7.13; p=0.007, 95% CI: -2.87 to -0.52). There was no difference in prescribing anti-DM in Hispanic vs. Non-Hispanic groups. Conclusions: In this study at two academic CSCs, there was no significant difference in prescribing anti-DM medications in Hispanic vs. Non-Hispanic groups, but insurance status may be associated with prescribing patterns. The provision of systematic care helped reduce healthcare disparity in AIS patients with newly diagnosed DM.

  PublisherDOI

• Renewable Electricity Management Cloud System for Smart Communities Using Advanced Machine Learning

  Energies · 2025-03-13 · 3 citations

  articleOpen access

  Based on the renewable energy assessment in 2023, it was found that only 21% of total electricity is generated using renewable sources. As the global demand for electricity rises in the AI world, the need for electricity management will increase and must be optimized. Based on research, many companies are working on green AI electricity management, but few companies are working on predicting shortages. To identify the rising electricity demand, predict the shortage, and to bring attention to consumption, this study focuses on the optimization of solar electricity generation, tracking its consumption, and forecasting the electricity shortages well in advance. This system demonstrates a novel approach using advanced machine learning, deep learning, and reinforcement learning to maximize solar energy utilization. This paper proposes and develops a community-based model that manages and analyzes multiple buildings’ energy usage, allowing the model to perform both distributed and aggregated decision-making, achieving an accuracy of 98.2% using stacking results of models with reinforcement learning. Concerning the real-world problem, this paper provides a sustainable solution by combining data-driven models with reinforcement learning, contributing to the current market need.

  PublisherOA PDFDOI

• Contention resolution with message deadlines

  Distributed Computing · 2025-07-12

  article1st authorCorresponding
  PublisherDOI

• Abstract TP298: Patterns of In-Hospital Prescribing of Post-Stroke Antidepressants in a Comprehensive Stroke Center

  Stroke · 2025-01-30

  article

  Background: Post stroke depression (PSD) affects over 33% of stroke survivors with the highest incidence occurring in the first year after stroke. Predictors of PSD include stroke severity, cognitive impairment, age, and physical disability. Antidepressant prescribing patterns during acute ischemic stroke (AIS) hospitalization may be inconsistent. This study evaluated antidepressant prescribing patterns at hospital discharge in acute AIS and hemorrhagic (ICH) stroke patients. Methods: We retrospectively examined prospectively collected data from an IRB-approved stroke registry at two academic Comprehensive Stroke Centers (CSC) between 1/1/2013 and 6/30/2024. We included patients with new diagnosis of AIS or ICH with motor hemiparesis. Patients with initial NIHSS 0 and baseline mRS >2 were excluded. A correlation matrix was constructed followed by stepwise linear regression. Logistic regression modeling was then used for any variables with p<0.10. Chi squared was used to examine nominal variables. Results: A total of 745 patients met study criteria, and 167 (22.4%) patients were prescribed antidepressants prior to discharge (AIS=137; ICH=30). Variables that were significantly correlated with prescribing antidepressants during stroke hospitalization were White race (p<0.001), length of stay (LOS) (p<0.001), discharge disposition to acute rehabilitation unit (p=0.02), history of smoking (p=0.002), history of substance abuse (p=0.006), history of depression (p<0.001), dementia (p=0.002), and patients with no medical history (p=0.02). Logistic regression model including LOS (p<0.001), history of depression (p=0.03), and no medical history (p=0.04) was significant in predicting prescribing antidepressants (p<0.001; r=0.30; r 2 =0.09). Conclusion: In patients hospitalized for AIS or hemorrhagic stroke with motor weakness, key demographic factors, LOS, and discharge to acute rehabilitation units were significant predictors for prescribing antidepressants prior to discharge. The independent predictors in our model account for 9% of the variance in prescribing antidepressants, suggesting that further studies are needed to understand other factors that may influence antidepressant prescribing patterns during stroke hospitalization.

  PublisherDOI

• Smart Green Energy Management for Campus: An Integrated Machine Learning and Reinforcement Learning Model

  Smart Cities · 2025-02-13 · 9 citations

  articleOpen access

  The increasing demand for energy efficiency and the integration of renewable energy sources have become crucial for sustainability in modern campuses. This work presents a smart green energy management system (SGEMS) that integrates a machine learning model and reinforcement learning (RL) to optimize energy consumption and solar generation across a green campus. Using historical data from three campus buildings, we developed a predictive model to forecast short-term energy consumption and solar generation. The XGBoost algorithm, combined with RL, demonstrated superior performance in predicting energy consumption and generation, outperforming other models with a root mean square error (RMSE) of 14.72, a mean absolute error (MAE) of 12.00, and a mean absolute percentage error (MAPE) of 2.18%. This work proposes a web-based interface for real-time energy monitoring and decision-making, helping users forecast power shortages and manage energy usage effectively. The proposed approach provides a scalable solution for campuses aiming to reduce reliance on external grids and increase energy efficiency, setting a benchmark for future green campus initiatives.

  PublisherOA PDFDOI

Recent grants

• CAREER: Provably Good Concurrency Platforms for Streaming Applications

  NSF · $456k · 2012–2018

• XPS: FP: Real-Time Scheduling of Parallel Tasks

  NSF · $766k · 2013–2019

• AF: SMALL: Collaborative Research: Data Structures for Parallel Algorithms

  NSF · $194k · 2012–2016

• XPS: FULL: FP: Collaborative Research: Taming parallelism: optimally exploiting high-throughput parallel architectures

  NSF · $330k · 2014–2020

• SPX: Collaborative Research: Eat your Wheaties: Multi-Grain Compilers for Parallel Builds at Every Scale

  NSF · $200k · 2017–2023

Frequent coauthors

• Brett C. Meyer

  University of California, San Diego

  45 shared

• Dawn M Meyer

  University of California, San Diego

  38 shared

• I-Ting Angelina Lee

  Washington University in St. Louis

  31 shared

• Royya Modir

  University of California, San Diego

  24 shared

• Thomas Hemmen

  University of California, San Diego

  22 shared

• Chenyang Lu

  19 shared

• Sanjoy Baruah

  Washington University in St. Louis

  17 shared

• Jeremy T. Fineman

  Georgetown University

  17 shared

Education

• Ph.D., Neurosciences

  University of California, San Diego

  2009

• M.S., Neurosciences

  University of California, San Diego

  2004

• B.S., Neuroscience

  University of California, San Diego

  2002