Search

Your search keyword '"Lee, Young M"' showing total 369 results
Start Over Author "Lee, Young M"
369 results on '"Lee, Young M"'

2. Neuroworsening in the Emergency Department Is a Predictor of Traumatic Brain Injury Intervention and Outcome: A TRACK-TBI Pilot Study.

Author

Yue, John K, Krishnan, Nishanth, Kanter, John H, Deng, Hansen, Okonkwo, David O, Puccio, Ava M, Madhok, Debbie Y, Belton, Patrick J, Lindquist, Britta E, Satris, Gabriela G, Lee, Young M, Umbach, Gray, Duhaime, Ann-Christine, Mukherjee, Pratik, Yuh, Esther L, Valadka, Alex B, DiGiorgio, Anthony M, Tarapore, Phiroz E, Huang, Michael C, Manley, Geoffrey T, and TRACK-TBI Investigators

Subjects
TRACK-TBI Investigators, Glasgow Coma Scale, emergency department, mortality, neurological examination, neuroworsening, patient outcome assessment, traumatic brain injury, Prevention, Physical Injury - Accidents and Adverse Effects, Clinical Research, Traumatic Head and Spine Injury, Brain Disorders, Traumatic Brain Injury (TBI), Neurosciences, Injuries and accidents, Good Health and Well Being, Clinical Sciences
Abstract

IntroductionNeuroworsening may be a sign of progressive brain injury and is a factor for treatment of traumatic brain injury (TBI) in intensive care settings. The implications of neuroworsening for clinical management and long-term sequelae of TBI in the emergency department (ED) require characterization.MethodsAdult TBI subjects from the prospective Transforming Research and Clinical Knowledge in Traumatic Brain Injury Pilot Study with ED admission and disposition Glasgow Coma Scale (GCS) scores were extracted. All patients received head computed tomography (CT) scan

Published
2023

3. Distributed Online Non-convex Optimization with Composite Regret

Author

Jiang, Zhanhong, Balu, Aditya, Lee, Xian Yeow, Lee, Young M., Hegde, Chinmay, and Sarkar, Soumik

Subjects
Computer Science - Machine Learning, Computer Science - Distributed, Parallel, and Cluster Computing, Statistics - Machine Learning
Abstract

Regret has been widely adopted as the metric of choice for evaluating the performance of online optimization algorithms for distributed, multi-agent systems. However, data/model variations associated with agents can significantly impact decisions and requires consensus among agents. Moreover, most existing works have focused on developing approaches for (either strongly or non-strongly) convex losses, and very few results have been obtained regarding regret bounds in distributed online optimization for general non-convex losses. To address these two issues, we propose a novel composite regret with a new network regret-based metric to evaluate distributed online optimization algorithms. We concretely define static and dynamic forms of the composite regret. By leveraging the dynamic form of our composite regret, we develop a consensus-based online normalized gradient (CONGD) approach for pseudo-convex losses, and it provably shows a sublinear behavior relating to a regularity term for the path variation of the optimizer. For general non-convex losses, we first shed light on the regret for the setting of distributed online non-convex learning based on recent advances such that no deterministic algorithm can achieve the sublinear regret. We then develop the distributed online non-convex optimization with composite regret (DINOCO) without access to the gradients, depending on an offline optimization oracle. DINOCO is shown to achieve sublinear regret; to our knowledge, this is the first regret bound for general distributed online non-convex learning., Comment: 41 pages, presented in allerton conference 2022

Published
2022

4. Asynchronous Training Schemes in Distributed Learning with Time Delay

Author

Wang, Haoxiang, Jiang, Zhanhong, Liu, Chao, Sarkar, Soumik, Jiang, Dongxiang, and Lee, Young M.

Subjects
Computer Science - Machine Learning
Abstract

In the context of distributed deep learning, the issue of stale weights or gradients could result in poor algorithmic performance. This issue is usually tackled by delay tolerant algorithms with some mild assumptions on the objective functions and step sizes. In this paper, we propose a different approach to develop a new algorithm, called $\textbf{P}$redicting $\textbf{C}$lipping $\textbf{A}$synchronous $\textbf{S}$tochastic $\textbf{G}$radient $\textbf{D}$escent (aka, PC-ASGD). Specifically, PC-ASGD has two steps - the $\textit{predicting step}$ leverages the gradient prediction using Taylor expansion to reduce the staleness of the outdated weights while the $\textit{clipping step}$ selectively drops the outdated weights to alleviate their negative effects. A tradeoff parameter is introduced to balance the effects between these two steps. Theoretically, we present the convergence rate considering the effects of delay of the proposed algorithm with constant step size when the smooth objective functions are weakly strongly-convex and nonconvex. One practical variant of PC-ASGD is also proposed by adopting a condition to help with the determination of the tradeoff parameter. For empirical validation, we demonstrate the performance of the algorithm with two deep neural network architectures on two benchmark datasets., Comment: This work has been submitted to the IEEE for possible publication

Published
2022

5. Paraparesis caused by intradural thoracic spinal granuloma secondary to organizing hematoma: illustrative case

Author

Yue, John K, Lee, Young M, Quintana, Daniel, Aabedi, Alexander A, Krishnan, Nishanth, Wozny, Thomas A, Andrews, John P, and Huang, Michael C

Subjects
Biomedical and Clinical Sciences, Clinical Sciences, Cardiovascular, inflammation, intradural hematoma, neurocritical care, spinal cord injury, spinal granuloma, spine surgery
Abstract

BackgroundSpinal granulomas form from infectious or noninfectious inflammatory processes and are rarely present intradurally. Intradural granulomas secondary to hematoma are unreported in the literature and present diagnostic and management challenges.ObservationsA 70-year-old man receiving aspirin presented with encephalopathy, subacute malaise, and right lower extremity weakness and was diagnosed with polysubstance withdrawal and refractory hypertension requiring extended treatment. Seven days after admission, he reported increased bilateral lower extremity (BLE) weakness. Magnetic resonance imaging showed T2-3 and T7-8 masses abutting the pia, with spinal cord compression at T2-3. He was transferred to the authors' institution, and work-up showed no vascular shunting or malignancy. He underwent T2-3 laminectomies for biopsy/resection. A firm, xanthochromic mass was resected en bloc. Pathology showed organizing hematoma without infection, vascular malformation, or malignancy. Subsequent coagulopathy work-up was unremarkable. His BLE strength significantly improved, and he declined resection of the inferior mass. He completed physical therapy and was cleared for placement in a skilled nursing facility.LessonsSpinal granulomas can mimic vascular lesions and malignancy. The authors present the first report of paraparesis caused by intradural granuloma secondary to organizing hematoma, preceded by severe refractory hypertension. Tissue diagnosis is critical, and resection is curative. These findings can inform the vigilant clinician for expeditious treatment.

Published
2022

6. MDPGT: Momentum-based Decentralized Policy Gradient Tracking

Author

Jiang, Zhanhong, Lee, Xian Yeow, Tan, Sin Yong, Tan, Kai Liang, Balu, Aditya, Lee, Young M., Hegde, Chinmay, and Sarkar, Soumik

Subjects
Computer Science - Machine Learning, Computer Science - Artificial Intelligence
Abstract

We propose a novel policy gradient method for multi-agent reinforcement learning, which leverages two different variance-reduction techniques and does not require large batches over iterations. Specifically, we propose a momentum-based decentralized policy gradient tracking (MDPGT) where a new momentum-based variance reduction technique is used to approximate the local policy gradient surrogate with importance sampling, and an intermediate parameter is adopted to track two consecutive policy gradient surrogates. Moreover, MDPGT provably achieves the best available sample complexity of $\mathcal{O}(N^{-1}\epsilon^{-3})$ for converging to an $\epsilon$-stationary point of the global average of $N$ local performance functions (possibly nonconcave). This outperforms the state-of-the-art sample complexity in decentralized model-free reinforcement learning, and when initialized with a single trajectory, the sample complexity matches those obtained by the existing decentralized policy gradient methods. We further validate the theoretical claim for the Gaussian policy function. When the required error tolerance $\epsilon$ is small enough, MDPGT leads to a linear speed up, which has been previously established in decentralized stochastic optimization, but not for reinforcement learning. Lastly, we provide empirical results on a multi-agent reinforcement learning benchmark environment to support our theoretical findings.

Published
2021

7. C2 translaminar screw fixation in pediatric occipitocervical fusion.

Author

Lee, Young M, Lu, Alex Y, Oh, Taemin, Hwang, Joan Y, Lu, Daniel C, and Sun, Peter P

Subjects
Cervical Vertebrae, Atlanto-Axial Joint, Humans, Kyphosis, Joint Instability, Treatment Outcome, Spinal Fusion, Retrospective Studies, Bone Screws, Child, Atlantoaxial fixation, Occipitocervical instability, Pediatric spine, Spinal fusion, Translaminar screw, Pediatric, Physical Injury - Accidents and Adverse Effects, Clinical Sciences, Neurology & Neurosurgery
Abstract

PurposeRigid occipitocervical (O-C) instrumentation can reduce the anterior pathology and has a high fusion rate in children with craniovertebral instability. Typically, axis (C2) screw fixation utilizes C1-C2 transarticular screws or C2 pars screws. However, anatomic variation may preclude these screw types due to the size of fixation elements or by placing the vertebral artery at risk for injury. Pediatric C2 translaminar screw fixation has low risk of vertebral artery injury and may be used when the anatomy is otherwise unsuitable for C1-C2 transarticular screws or C2 pars screws.MethodsWe retrospectively reviewed a neurosurgical database at UCSF Benioff Children's Hospital Oakland for patients who had undergone a cervical spinal fusion that utilized translaminar screws for occipitocervical instrumentation between 2002 and 2020. We then reviewed the operative records to determine the parameters of C2 screw fixations performed. Demographic and all other relevant clinical data were then recorded.ResultsTwenty-five patients ranging from 2 to 18 years of age underwent O-C fusion, with a total of 43 translaminar screws at C2 placed. Twenty-three patients were fused (92%) after initial surgery with a mean follow-up of 43 months. Two patients, both with Down syndrome, had a nonunion. Another 2 patients had a superficial wound dehiscence that required wound revision. One patient died of unknown cause 7 months after surgery. One patient developed an adjacent-level kyphosis.ConclusionWhen performing occipitocervical instrumentation in the pediatric population, C2 translaminar screw fixation is an effective option to other methods of C2 screw fixation dependent on anatomic feasibility.

Published
2022

8. Retrosigmoid Craniectomy with a Layered Soft Tissue Dissection and Hydroxyapatite Reconstruction: Technical Note, Surgical Video, Regional Anatomy, and Outcomes

Author

Magill, Stephen T, Lee, Young M, Rubio, Roberto R, Nguyen, Minh P, Heilman, Carl B, and McDermott, Michael W

Subjects
Biomedical and Clinical Sciences, Clinical Sciences, retrosigmoid, cerebellopontine angle, anatomy, neuroanatomy, CSF leak, suboccipital, retromastoid, hydroxyapatite, reconstruction, Neurology & Neurosurgery, Dentistry
Abstract

Introduction  There are many reported modifications to the retrosigmoid approach including variations in skin incisions, soft tissue dissection, bone removal/replacement, and closure. Objective  The aim of this study was to report the technical nuances developed by two senior skull base surgeons for retrosigmoid craniectomy with reconstruction and provide anatomic dissections, surgical video, and outcomes. Methods  The regional soft tissue and bony anatomy as well as the steps for our retrosigmoid craniectomy were recorded with photographs, anatomic dissections, and video. Records from 2017 to 2019 were reviewed to determine the incidence of complications after the authors began using the described approach. Results  Dissections of the relevant soft tissue, vascular, and bony structures were performed. Key surgical steps are (1) a retroauricular C-shaped skin incision, (2) developing a skin and subgaleal tissue flap of equal thickness above the fascia over the temporalis and sub-occipital muscles, (3) creation of subperiosteal soft tissue planes over the top of the mastoid and along the superior nuchal line to expose the suboccipital region, (4) closure of the craniectomy defect with in-lay titanium mesh and overlay hydroxyapatite cranioplasty, and (5) reapproximation of the soft tissue edges during closure. Complications in 40 cases were pseudomeningocele requiring shunt ( n  = 3, 7.5%), wound infection ( n  = 1, 2.5%), and aseptic meningitis ( n  = 1, 2.5%). There were no incisional cerebrospinal fluid leaks. Conclusion  The relevant regional anatomy and a revised technique for retrosigmoid craniectomy with reconstruction have been presented with acceptable results. Readers can consider this technique when using the retrosigmoid approach for pathology in the cerebellopontine angle.

Published
2022

10. Decentralized Deep Learning using Momentum-Accelerated Consensus

Author

Balu, Aditya, Jiang, Zhanhong, Tan, Sin Yong, Hedge, Chinmay, Lee, Young M, and Sarkar, Soumik

Subjects
Computer Science - Machine Learning, Computer Science - Distributed, Parallel, and Cluster Computing, Statistics - Machine Learning
Abstract

We consider the problem of decentralized deep learning where multiple agents collaborate to learn from a distributed dataset. While there exist several decentralized deep learning approaches, the majority consider a central parameter-server topology for aggregating the model parameters from the agents. However, such a topology may be inapplicable in networked systems such as ad-hoc mobile networks, field robotics, and power network systems where direct communication with the central parameter server may be inefficient. In this context, we propose and analyze a novel decentralized deep learning algorithm where the agents interact over a fixed communication topology (without a central server). Our algorithm is based on the heavy-ball acceleration method used in gradient-based optimization. We propose a novel consensus protocol where each agent shares with its neighbors its model parameters as well as gradient-momentum values during the optimization process. We consider both strongly convex and non-convex objective functions and theoretically analyze our algorithm's performance. We present several empirical comparisons with competing decentralized learning methods to demonstrate the efficacy of our approach under different communication topologies.

Published
2020

12. Deep Transfer Learning for Thermal Dynamics Modeling in Smart Buildings

Author

Jiang, Zhanhong and Lee, Young M.

Subjects
Statistics - Machine Learning, Computer Science - Machine Learning, Electrical Engineering and Systems Science - Systems and Control
Abstract

Thermal dynamics modeling has been a critical issue in building heating, ventilation, and air-conditioning (HVAC) systems, which can significantly affect the control and maintenance strategies. Due to the uniqueness of each specific building, traditional thermal dynamics modeling approaches heavily depending on physics knowledge cannot generalize well. This study proposes a deep supervised domain adaptation (DSDA) method for thermal dynamics modeling of building indoor temperature evolution and energy consumption. A long short term memory network based Sequence to Sequence scheme is pre-trained based on a large amount of data collected from a building and then adapted to another building which has a limited amount of data by applying the model fine-tuning. We use four publicly available datasets: SML and AHU for temperature evolution, long-term datasets from two different commercial buildings, termed as Building 1 and Building 2 for energy consumption. We show that the deep supervised domain adaptation is effective to adapt the pre-trained model from one building to another building and has better predictive performance than learning from scratch with only a limited amount of data., Comment: 5 pages, 2 figures; Accepted at 2019 IEEE International Conference on Big Data (IEEE BigData 2019)

Published
2019

13. On Higher-order Moments in Adam

Author

Jiang, Zhanhong, Balu, Aditya, Tan, Sin Yong, Lee, Young M, Hegde, Chinmay, and Sarkar, Soumik

Subjects
Computer Science - Machine Learning, Statistics - Machine Learning
Abstract

In this paper, we investigate the popular deep learning optimization routine, Adam, from the perspective of statistical moments. While Adam is an adaptive lower-order moment based (of the stochastic gradient) method, we propose an extension namely, HAdam, which uses higher order moments of the stochastic gradient. Our analysis and experiments reveal that certain higher-order moments of the stochastic gradient are able to achieve better performance compared to the vanilla Adam algorithm. We also provide some analysis of HAdam related to odd and even moments to explain some intriguing and seemingly non-intuitive empirical results., Comment: Accepted in Beyond First Order Methods in Machine Learning workshop in 33rd Conference on Neural Information Processing Systems (NeurIPS 2019), Vancouver, Canada

Published
2019

14. Cerebrospinal Fluid Leaks and Pseudomeningocele after Posterior Fossa Surgery: Effect of an Autospray Dural Sealant.

Author

Lee, Young M, Ordaz, Angel, Durcanova, Beata, Viner, Jennifer A, Theodosopoulos, Philip V, Aghi, Manish K, and McDermott, Michael W

Subjects
cost analysis, dural sealant, posterior fossa surgery, pseudomeningocele, Medical and Health Sciences
Abstract

Background Posterior fossa craniotomies can be complicated by cerebrospinal fluid (CSF) leaks, infection, meningitis, neurologic deficits, and intracranial hypotension caused by defective closure of the dura. Secondary dural closures such as pericranial graft, muscle graft, glue, sealants, or fat graft are used. However, there have been few studies examining the use of sealants with a polyethylene glycol and polyethylenimine component. Objective We studied the effect of one such sealant, Adherus® (HyperBranch Medical Technology, Durham, NC, USA), as an adjunct to secondary closure methods in the reduction of the use of abdominal fat grafting and lumbar puncture/drains. Methods We retrospectively reviewed the surgical records of all patients undergoing posterior fossa cranial surgery during a two-year period at a tertiary university affiliated medical center. Results Overall, data a total of 122 patients (62 in the no Adherus and 60 in the Adherus group) were collected. There was no statistically significant difference in the 30-day incisional CSF leak rate (4.1% vs. 6.5%; p=0.183), 30-day non-incisional CSF leak rate (11.3% vs. 5.0%; p=0.205), and 30-day pseudomeningocele rate (16.1% vs. 13.3%; p=0.663) in the no Adherus and Adherus groups, respectively. However, there was a significant reduction in the use of abdominal fat grafting (0% vs. 30.7%; p

Published
2020

15. Polytrauma Is Associated with Increased Three- and Six-Month Disability after Traumatic Brain Injury: A TRACK-TBI Pilot Study.

Author

Yue, John K, Satris, Gabriela G, Dalle Ore, Cecilia L, Huie, J Russell, Deng, Hansen, Winkler, Ethan A, Lee, Young M, Vassar, Mary J, Taylor, Sabrina R, Schnyer, David M, Lingsma, Hester F, Puccio, Ava M, Yuh, Esther L, Mukherjee, Pratik, Valadka, Alex B, Ferguson, Adam R, Markowitz, Amy J, Okonkwo, David O, and Manley, Geoffrey T

Subjects
disability, functional outcome, outcome measure, polytrauma, traumatic brain injury, Neurosciences, Injury - Traumatic brain injury, Clinical Research, Injury - Trauma - (Head and Spine), Injury (total) Accidents/Adverse Effects, Brain Disorders, Rehabilitation, Injuries and accidents
Abstract

Polytrauma and traumatic brain injury (TBI) frequently co-occur and outcomes are routinely measured by the Glasgow Outcome Scale-Extended (GOSE). Polytrauma may confound GOSE measurement of TBI-specific outcomes. Adult patients with TBI from the prospective Transforming Research and Clinical Knowledge in Traumatic Brain Injury Pilot (TRACK-TBI Pilot) study had presented to a Level 1 trauma center after injury, received head computed tomography (CT) within 24 h, and completed the GOSE at 3 months and 6 months post-injury. Polytrauma was defined as an Abbreviated Injury Score (AIS) ≥3 in any extracranial region. Univariate regressions were performed using known GOSE clinical cutoffs. Multi-variable regressions were performed for the 3- and 6-month GOSE, controlling for known demographic and injury predictors. Of 361 subjects (age 44.9 ± 18.9 years, 69.8% male), 69 (19.1%) suffered polytrauma. By Glasgow Coma Scale (GCS) assessment, 80.1% had mild, 5.8% moderate, and 14.1% severe TBI. On univariate logistic regression, polytrauma was associated with increased odds of moderate disability or worse (GOSE ≤6; 3 month odds ratio [OR] = 2.57 [95% confidence interval (CI): 1.50-4.41; 6 month OR = 1.70 [95% CI: 1.01-2.88]) and death/severe disability (GOSE ≤4; 3 month OR = 3.80 [95% CI: 2.03-7.11]; 6 month OR = 3.33 [95% CI: 1.71-6.46]). Compared with patients with isolated TBI, more polytrauma patients experienced a decline in GOSE from 3 to 6 months (37.7 vs. 24.7%), and fewer improved (11.6 vs. 22.6%). Polytrauma was associated with greater univariate ordinal odds for poorer GOSE (3 month OR = 2.79 [95% CI: 1.73-4.49]; 6 month OR = 1.73 [95% CI: 1.07-2.79]), which was conserved on multi-variable ordinal regression (3 month OR = 3.05 [95% CI: 1.76-5.26]; 6 month OR = 2.04 [95% CI: 1.18-3.42]). Patients with TBI with polytrauma are at greater risk for 3- and 6-month disability compared with those with isolated TBI. Methodological improvements in assessing TBI-specific disability, versus disability attributable to all systemic injuries, will generate better TBI outcomes assessment tools.

Published
2020

18. Schwannomatosis of the Spinal Accessory Nerve: A Case Report

Author

Morshed, Ramin A, Lee, Anthony T, Lee, Young M, Chin, Cynthia T, and Jacques, Line

Subjects
Biomedical and Clinical Sciences, Neurosciences, Clinical Sciences, Human Genome, Neurofibromatosis, Prevention, Genetics, Pediatric, Rare Diseases, Aetiology, 2.1 Biological and endogenous factors, Neurological, spinal accessory nerve tumor, schwannomatosis, schwannoma, neck mass, diffusion tensor imaging, Neurology & Neurosurgery, Clinical sciences, Allied health and rehabilitation science
Abstract

Schwannomatosis is a distinct syndrome characterized by multiple peripheral nerve schwannomas that can be sporadic or familial in nature. Cases affecting the lower cranial nerves are infrequent. Here, the authors present a rare case of schwannomatosis affecting the left spinal accessory nerve. Upon genetic screening, an in-frame insertion at codon p.R177 of the Sox 10 gene was observed. There were no identifiable alterations in NF1, NF2, LZTR1, and SMARCB1. This case demonstrates a rare clinical presentation of schwannomatosis in addition to a genetic aberration that has not been previously reported in this disease context.

Published
2019

22. Performance of the IMPACT and CRASH prognostic models for traumatic brain injury in a contemporary multicenter cohort: a TRACK-TBI study

Author

Yue, John K., primary, Lee, Young M., additional, Sun, Xiaoying, additional, van Essen, Thomas A., additional, Elguindy, Mahmoud M., additional, Belton, Patrick J., additional, Pisică, Dana, additional, Mikolic, Ana, additional, Deng, Hansen, additional, Kanter, John H., additional, McCrea, Michael A., additional, Bodien, Yelena G., additional, Satris, Gabriela G., additional, Wong, Justin C., additional, Ambati, Vardhaan S., additional, Grandhi, Ramesh, additional, Puccio, Ava M., additional, Mukherjee, Pratik, additional, Valadka, Alex B., additional, Tarapore, Phiroz E., additional, Huang, Michael C., additional, DiGiorgio, Anthony M., additional, Markowitz, Amy J., additional, Yuh, Esther L., additional, Okonkwo, David O., additional, Steyerberg, Ewout W., additional, Lingsma, Hester F., additional, Menon, David K., additional, Maas, Andrew I. R., additional, Jain, Sonia, additional, Manley, Geoffrey T., additional, _, _, additional, Badjatia, Neeraj, additional, Barber, Jason, additional, Chesnut, Randall M., additional, Diaz-Arrastia, Ramon, additional, Duhaime, Ann-Christine, additional, Eagle, Shawn R., additional, Etemad, Leila L., additional, Fabian, Brian, additional, Ferguson, Adam R., additional, Foreman, Brandon, additional, Gardner, Raquel C., additional, Giacino, Joseph T., additional, Gopinath, Shankar, additional, Gotthardt, Christine J., additional, Hamidi, Sabah, additional, Huie, J. Russell, additional, Keene, C. Dirk, additional, Korley, Frederick K., additional, Madhok, Debbie Y., additional, Madden, Christopher, additional, Merchant, Randall, additional, Nelson, Lindsay D., additional, Ngwenya, Laura B., additional, Robertson, Claudia S., additional, Rodgers, Richard B, additional, Schneider, Andrea L. C., additional, Schnyer, David M., additional, Stein, Murray B., additional, Taylor, Sabrina R., additional, Temkin, Nancy R., additional, Torres-Espin, Abel, additional, Tracey, Joye X., additional, Vassar, Mary J., additional, Wang, Kevin K. W., additional, and Zafonte, Ross D., additional

Published
2024
Full Text
View/download PDF

28. Descriptions of Motor Vehicle Collisions by Participants in Emergency Department–Based Studies: Are They Accurate?

Author

Lee, Young M, Platts-Mills, Timothy F, MacWilliams, Joel B, Sochor, Mark R, Jones, Jeffrey S, Domeier, Robert M, Schneider, Lawrence W, and McLean, Samuel A

Subjects
motor vehicle crash, research methodology, seat belt, air bag, police officer, Biomechanics
Abstract

Introduction: We examined the accuracy of research participant characterizations of motor vehicle collisions (MVC).Methods: We conducted an emergency department-based prospective study of adults presenting for care after experiencing an MVC. Study participants completed a structured clinical interview that assessed the number of lanes of the road where the collision took place, vehicle type, road condition, speed limit, seat belt use, airbag deployment, vehicle damage, time of collision, and use of ambulance transportation. Study participant data were then compared with information recorded by Michigan State Police at the scene of the MVC. Agreement between research participant reports and police-reported data were assessed by using percentage agreement and j coefficients for categorical variables and correlation coefficients for continuous variables.Results: There were 97 study participants for whom emergency department interviews and Michigan State Police Report information were available. Percentage agreement was 51% for number of lanes,76% for car drivability, 88% for road condition, 91% for vehicle type, 92% for seat belt use, 94% for airbag deployment, 96% for speed limit, 97% for transportation by ambulance, and 99% for vehicle seat position. j values were 0.32 for seat belt use, 0.34 for number of lanes, 0.73 for vehicle type, 0.76 for speed limit, 0.77 for road condition, 0.87 for airbag deployment, 0.90 for vehicle seat position, and 0.94for transport by ambulance. Correlation coefficients were 0.95 for the time of the collision, and 0.58 for extent of damage to the vehicle. Most discrepancies between patients and police about extent of vehicle damage occurred for cases in which the patient reported moderate or severe damage but the police reported only slight damage.Conclusion: For most MVC characteristics, information reported by research participants was consistent with police-reported data. Agreement was moderate or high for characteristics of greatestrelevance to injury biomechanics. These results suggest that research participant report is an acceptable source of collision information. [West J Emerg Med. 2012;13(4):329–334.]

Published
2012

30. MDPGT: Momentum-Based Decentralized Policy Gradient Tracking

Author

Jiang, Zhanhong, Lee, Xian Yeow, Tan, Sin Yong, Tan, Kai Liang, Balu, Aditya, Lee, Young M., Hegde, Chinmay, and Sarkar, Soumik

Subjects
FOS: Computer and information sciences, Computer Science - Machine Learning, Artificial Intelligence (cs.AI), Computer Science - Artificial Intelligence, General Medicine, Machine Learning (cs.LG)
Abstract

We propose a novel policy gradient method for multi-agent reinforcement learning, which leverages two different variance-reduction techniques and does not require large batches over iterations. Specifically, we propose a momentum-based decentralized policy gradient tracking (MDPGT) where a new momentum-based variance reduction technique is used to approximate the local policy gradient surrogate with importance sampling, and an intermediate parameter is adopted to track two consecutive policy gradient surrogates. Moreover, MDPGT provably achieves the best available sample complexity of $\mathcal{O}(N^{-1}\epsilon^{-3})$ for converging to an $\epsilon$-stationary point of the global average of $N$ local performance functions (possibly nonconcave). This outperforms the state-of-the-art sample complexity in decentralized model-free reinforcement learning, and when initialized with a single trajectory, the sample complexity matches those obtained by the existing decentralized policy gradient methods. We further validate the theoretical claim for the Gaussian policy function. When the required error tolerance $\epsilon$ is small enough, MDPGT leads to a linear speed up, which has been previously established in decentralized stochastic optimization, but not for reinforcement learning. Lastly, we provide empirical results on a multi-agent reinforcement learning benchmark environment to support our theoretical findings.

Published
2022

31. Data-driven control of airborne infection risk and energy use in buildings

Author

Risbeck, Michael James, primary, Cohen, Alexander Ethan, additional, Douglas, Jonathan D, additional, Jiang, Zhanhong, additional, Fanone, Carlo M, additional, Bowes, Karen, additional, Doughty, Jim, additional, Turnbull, Martin, additional, DiBerardinis, Louis, additional, Lee, Young M, additional, and Bazant, Martin Z., additional

Published
2023
Full Text
View/download PDF

43. Assessment of Airborne Disease Transmission Risk and Energy Impact of HVAC Mitigation Strategies

Author

Martin Z. Bazant, Michael J. Risbeck, Zhanhong Jiang, Jonathan D. Douglas, Lee Young M, and Kirk H. Drees

Subjects
business.industry, Airflow, Energy impact, medicine.disease, Airborne disease, Energy analysis, law.invention, Transmission (telecommunications), Risk analysis (engineering), law, HVAC, Ventilation (architecture), medicine, Environmental science, Significant risk, business
Abstract

The COVID-19 pandemic has focused renewed attention on the ways in which building HVAC systems may be operated to mitigate the risk of airborne disease transmission. The most common suggestion is to increase outdoor-air ventilation rates so as to dilute the concentrations of infectious aerosol particles indoors. Although this strategy does reduce the likelihood of disease spread, it is often much more costly than other strategies that provide equivalent particle removal or deactivation. To address this tradeoff and arrive at practical recommendations, we explain how different mitigation strategies can be expressed in terms of equivalent outdoor air (EOA) to provide a common basis for energy analysis. We then show the effects of each strategy on EOA delivery and energy cost in simulations of realistic buildings in a variety of climates. Key findings are that in-duct filtration is often the most efficient mitigation strategy, while significant risk reduction generally requires increasing total airflow to the system, either through adjusted HVAC setpoints or standalone disinfection devices.

Published
2021

44. Modeling and Multiobjective Optimization of Indoor Airborne Disease Transmission Risk and Associated Energy Consumption for Building HVAC Systems

Author

Zhanhong Jiang, Michael J. Risbeck, Jonathan D. Douglas, Lee Young M, Kirk H. Drees, and Martin Z. Bazant

Subjects
Flexibility (engineering), business.industry, Mechanical Engineering, HVAC control system, Building and Construction, Energy consumption, medicine.disease, Multi-objective optimization, Airborne transmission, Airborne disease, Article, Reliability engineering, Control theory, HVAC, medicine, Electrical and Electronic Engineering, business, Civil and Structural Engineering
Abstract

The COVID-19 pandemic has renewed interest in assessing how the operation of HVAC systems influences the risk of airborne disease transmission in buildings. Various processes, such as ventilation and filtration, have been shown to reduce the probability of disease spread by removing or deactivating exhaled aerosols that potentially contain infectious material. However, such qualitative recommendations fail to specify how much of these or other disinfection techniques are needed to achieve acceptable risk levels in a particular space. An additional complication is that application of these techniques inevitably increases energy costs, the magnitude of which can vary significantly based on local weather. Moreover, the operational flexibility available to the HVAC system may be inherently limited by equipment capacities and occupant comfort requirements. Given this knowledge gap, we propose a set of dynamical models that can be used to estimate airborne transmission risk and energy consumption for building HVAC systems based on controller setpoints and a forecast of weather conditions. By combining physics-based material balances with phenomenological models of the HVAC control system, it is possible to predict time-varying airflows and other HVAC variables, which are then used to calculate key metrics. Through a variety of examples involving real and simulated commercial buildings, we show that our models can be used for monitoring purposes by applying them directly to transient building data as operated, or they may be embedded within a multi-objective optimization framework to evaluate the tradeoff between infection risk and energy consumption. By combining these applications, building managers can determine which spaces are in need of infection risk reduction and how to provide that reduction at the lowest energy cost. The key finding is that both the baseline infection risk and the most energy-efficient disinfection strategy can vary significantly from space to space and depend sensitively on the weather, thus underscoring the importance of the quantitative predictions provided by the models.

Published
2021

45. Quantifying the Tradeoff Between Energy Consumption and the Risk of Airborne Disease Transmission for Building HVAC Systems

Author

Zhanhong Jiang, Martin Z. Bazant, Michael J. Risbeck, Lee Young M, Kirk H. Drees, and Jonathan D. Douglas

Subjects
Fluid Flow and Transfer Processes, Environmental Engineering, business.industry, Computer science, fungi, food and beverages, Building and Construction, Energy consumption, medicine.disease, Airborne transmission, Airborne disease, Automotive engineering, law.invention, Transmission (telecommunications), Risk analysis (engineering), law, Ventilation (architecture), HVAC, Sustainability, medicine, Environmental science, business, Baseline (configuration management), Filtration, Energy (signal processing)
Abstract

Since the advent of the COVID-19 pandemic, there has been renewed interest in determining how the operation of building HVAC systems influences the risk of airborne transmission of disease. It has been established that combinations of increased ventilation, improved filtration, and other disinfection techniques can reduce the likelihood of transmission by removing or deactivating the airborne particles that potentially contain infectious material. However, when such guidance is general and qualitative in nature, it is extremely difficult for building managers to make informed decisions, as there is no quantitative information about how much risk reduction is provided. Furthermore, the actions that could be taken almost always require additional energy consumption by the HVAC system, and so in the absence of building-specific analysis, it is possible that chosen strategies might simply be wasting energy without providing meaningful reduction in transmission risk. To address this knowledge gap, we propose simplified steady-state models that can be used to quantify both the expected infection rate and the associated HVAC energy consumption that result from baseline operation and hypothetical changes. The transmission rate is modeled by considering the airborne concentration of infectious particles that would result from the activity-dependent respiration of an infector in the space, the physical dimensions of the space, and operation of the HVAC system. By formulating all disinfection mechanisms in terms of “equivalent outdoor air”, a common basis is established for comparing and combining different strategies. Energy consumption can then be estimated by considering the change in HVAC variables (e.g., flow rates and temperatures) and applying standard models. To illustrate the insights provided by these models, we present examples of how the proposed analysis can be applied to specific spaces, highlighting the fact that underlying transmission risk (and thus also the energy-optimal disinfection strategies) can vary significantly from building to building and even from space to space within the same building. The overall goal is to empower building managers to fully assess the tradeoff between energy consumption and infection risk so that they can more effectively target their disinfection efforts and take actions that are consistent with current health and sustainability priorities.

Published
2021

Catalog

Books, media, physical & digital resources
See catalog results