Peter Hu
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Peter Hu has a total of 78 co-authors. Here are the top co-authors by number of shared publications:
- Shiming Yang
Department of Anesthesiology, University of Maryland School of Medicine, Baltimore, MD
6 shared publications - Chein-I Chang
Center for Hyperspectral Imaging in Remote Sensing (CHIRS), Information and Technology College, Dalian Maritime University, Dalian, China
5 shared publications - Neeraj Badjatia
Program in Trauma, University of Maryland School of Medicine, Baltimore, USA
3 shared publications - Samuel M. Galvagno
University of Maryland School of Medicine
3 shared publications - Thomas Michael Scalea
R Adams Cowley Shock Trauma Center, University of Maryland School of Medicine, Baltimore, USA
3 shared publications - William Teeter
Emergency Medicine, University of Maryland Baltimore School of Medicine, Baltimore, Maryland, USA
3 shared publications
Latest Publications
- 2 February 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Geoscience and Remote Sensing
- Vol. 64 (01962892) , 1-22
Cascaded Spectral–Spatial Hyperspectral Anomaly Detection
An Iterative spectral-spatial hyperspectral anomaly detection (ISSHAD) was recently developed to augment the data cube to be processed to construct a new data cube via a feedback process in an iterative manner. As a result of such a feedback process hyperspectral anomaly detection (ISSHAD) significantly improve HAD performance. As an opposite approach to ISSHAD, this paper develops a new approach to HAD, called cascaded spectral-spatial hyperspectral anomaly detection (CSSHAD), which performs HAD by a cascade of anomaly detection modules (ADMs) in a feedforward network in which each cascaded anomaly detection module (ADM) implements the same anomaly detector to produce an anomaly detection map (ADMap). Then a spatial filter is designed to capture spatial information from the ADMap to generate a real-valued spatial filtered ADMap (SF-Map) which will be used as the input to the next following ADM. Finally, the depth of CSSHAD is determined by an automatic stopping rule. To further improve anomaly detectability and background suppressibility, a recent developed concept, effective anomaly space (EAS) is incorporated into CSSHAD to derive EAS-CSSHAD. Unlike ISSHAD which is a feedback HAD network, CSSHAD is a feedforward HAD network that operates a cascade of ADMs by feeding forward SFMaps in a similar manner that ISSHAD uses an iterative feedback process. Accordingly, CSSHAD and ISSHAD can be considered to dual HAD methods as companion methods. Experimental results also demonstrate that CSSHAD performs comparably to ISSHAD with less computational complexity due to its use of feedforward structure rather than feedback process.
- 2 February 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Geoscience and Remote Sensing
- Vol. 64 (01962892) , 1-22
Cascaded Spectral–Spatial Hyperspectral Anomaly Detection
An Iterative spectral-spatial hyperspectral anomaly detection (ISSHAD) was recently developed to augment the data cube to be processed to construct a new data cube via a feedback process in an iterative manner. As a result of such a feedback process hyperspectral anomaly detection (ISSHAD) significantly improve HAD performance. As an opposite approach to ISSHAD, this paper develops a new approach to HAD, called cascaded spectral-spatial hyperspectral anomaly detection (CSSHAD), which performs HAD by a cascade of anomaly detection modules (ADMs) in a feedforward network in which each cascaded anomaly detection module (ADM) implements the same anomaly detector to produce an anomaly detection map (ADMap). Then a spatial filter is designed to capture spatial information from the ADMap to generate a real-valued spatial filtered ADMap (SF-Map) which will be used as the input to the next following ADM. Finally, the depth of CSSHAD is determined by an automatic stopping rule. To further improve anomaly detectability and background suppressibility, a recent developed concept, effective anomaly space (EAS) is incorporated into CSSHAD to derive EAS-CSSHAD. Unlike ISSHAD which is a feedback HAD network, CSSHAD is a feedforward HAD network that operates a cascade of ADMs by feeding forward SFMaps in a similar manner that ISSHAD uses an iterative feedback process. Accordingly, CSSHAD and ISSHAD can be considered to dual HAD methods as companion methods. Experimental results also demonstrate that CSSHAD performs comparably to ISSHAD with less computational complexity due to its use of feedforward structure rather than feedback process.
- Open Access
- 22 September 2025
- Published by Wolters Kluwer Health in Critical Care Explorations
- Vol. 7 (10) , e1319
Predictors of Anemia Intolerance for Real-Time Transfusion Decision-Making During Resuscitation of Trauma Subjects: A Machine Learning Approach Using Heart Rate Variability
OBJECTIVES: RBC transfusion in anemic patients with sustainable tolerance may cause harm, emphasizing the need for reliable metrics that quantify adequacy (oxygen delivery ≥ demand) and sustainability (oxygen delivery remains adequate without transfusion) of compensatory physiology. Our objective was to identify personalized predictors of anemia intolerance (inadequate and unsustainable physiologic compensation) that predict the likelihood of transfusion benefit. We studied adult trauma subjects at arrival to the emergency department, employing machine learning to evaluate ability of heart rate variability (HRV) to predict subsequent need for clinically indicated significant RBC transfusion. DESIGN: This single-center retrospective cohort study used electronic medical records data from patients admitted to a specialized trauma care hospital between January 2016 and December 2018. SETTING: Trauma resuscitation unit (TRU). PATIENTS: Adult trauma subjects with at least 3 hours of stay in the TRU, without RBC transfusion during the first hour at TRU but, with receipt or nonreceipt of transfusion in the second and/or third hour were included. Availability of electrocardiogram tracings for at least 50% of the first hour of stay in the TRU was also considered for inclusion in the study. INTERVENTIONS: None. MEASUREMENTS AND MAIN RESULTS: The primary binary outcome variable, a clinically indicated significant transfusion, was if a subject received RBC transfusion or not during the second and third hour stay in the TRU (transfusion vs. no transfusion). Patient clinical information, and HRV parameters generated from a 5-minute electrocardiogram recording during the first hour of admission were used as predictors for predicting transfusion. We evaluated five predefined prediction models for transfusion using random forest algorithm, varying the inclusion of demographic, clinical, trauma, and HRV variables. Model predictive performance was assessed using area under the receiver operating characteristic curve (AUROC), sensitivity, specificity, and Shapley analysis was conducted to identify key contributing variables. The analysis included 269 patients (126: transfusion cohort and 133: no transfusion cohort), who met the inclusion criteria. The model, which included demographic, clinical laboratory, trauma, and HRV variables, had an AUROC of 0.86, a sensitivity of 78%, and a specificity of 75% in predicting transfusion throughout the 3-hour study period. The model with only HRV variables showed comparable predictive performance (AUROC: 0.72) compared with other models with less than 35% false positive and negative rates. Among HRV parameters, lower values of log-transformed very low frequency absolute power predicted transfusion consistently. CONCLUSIONS: HRV parameters collected during the first 5–10 minutes after admission, when combined with basic clinical information that is immediately available upon emergency admission, augmented ability to predict potential for RBC transfusion, suggesting this metric may be incorporated into structured approaches to personalized transfusion decision-making.
- 22 September 2025
- Published by Wolters Kluwer Health in Critical Care Explorations
- Vol. 7 (10) , e1319
Open AccessPredictors of Anemia Intolerance for Real-Time Transfusion Decision-Making During Resuscitation of Trauma Subjects: A Machine Learning Approach Using Heart Rate Variability
OBJECTIVES: RBC transfusion in anemic patients with sustainable tolerance may cause harm, emphasizing the need for reliable metrics that quantify adequacy (oxygen delivery ≥ demand) and sustainability (oxygen delivery remains adequate without transfusion) of compensatory physiology. Our objective was to identify personalized predictors of anemia intolerance (inadequate and unsustainable physiologic compensation) that predict the likelihood of transfusion benefit. We studied adult trauma subjects at arrival to the emergency department, employing machine learning to evaluate ability of heart rate variability (HRV) to predict subsequent need for clinically indicated significant RBC transfusion. DESIGN: This single-center retrospective cohort study used electronic medical records data from patients admitted to a specialized trauma care hospital between January 2016 and December 2018. SETTING: Trauma resuscitation unit (TRU). PATIENTS: Adult trauma subjects with at least 3 hours of stay in the TRU, without RBC transfusion during the first hour at TRU but, with receipt or nonreceipt of transfusion in the second and/or third hour were included. Availability of electrocardiogram tracings for at least 50% of the first hour of stay in the TRU was also considered for inclusion in the study. INTERVENTIONS: None. MEASUREMENTS AND MAIN RESULTS: The primary binary outcome variable, a clinically indicated significant transfusion, was if a subject received RBC transfusion or not during the second and third hour stay in the TRU (transfusion vs. no transfusion). Patient clinical information, and HRV parameters generated from a 5-minute electrocardiogram recording during the first hour of admission were used as predictors for predicting transfusion. We evaluated five predefined prediction models for transfusion using random forest algorithm, varying the inclusion of demographic, clinical, trauma, and HRV variables. Model predictive performance was assessed using area under the receiver operating characteristic curve (AUROC), sensitivity, specificity, and Shapley analysis was conducted to identify key contributing variables. The analysis included 269 patients (126: transfusion cohort and 133: no transfusion cohort), who met the inclusion criteria. The model, which included demographic, clinical laboratory, trauma, and HRV variables, had an AUROC of 0.86, a sensitivity of 78%, and a specificity of 75% in predicting transfusion throughout the 3-hour study period. The model with only HRV variables showed comparable predictive performance (AUROC: 0.72) compared with other models with less than 35% false positive and negative rates. Among HRV parameters, lower values of log-transformed very low frequency absolute power predicted transfusion consistently. CONCLUSIONS: HRV parameters collected during the first 5–10 minutes after admission, when combined with basic clinical information that is immediately available upon emergency admission, augmented ability to predict potential for RBC transfusion, suggesting this metric may be incorporated into structured approaches to personalized transfusion decision-making.
- 14 April 2025
- Published by SAGE Publications in Tissue Engineering Part A
- Vol. 31 (7-8) , 285-302
From Promise to Practice: Recent Growth in 30 Years of Tissue Engineering Commercialization
This perspective, marking the 30th anniversary of the Tissue Engineering journal, discusses the exciting trends in the global commercialization of tissue engineering technology. Within a historical context, we present an evolution of challenges and a discussion of the last 5 years of global commercial successes and emerging market trends, highlighting the continued expansion of the field in the northeastern United States. This leads to an overview of the last 5 years’ progress in clinical trials for tissue-engineered therapeutics, including an analysis of trends in success and failure. Finally, we provide a broad overview of preclinical research and a perspective on where the state-of-the-art lies on the horizon.
- 14 April 2025
- Published by SAGE Publications in Tissue Engineering Part A
- Vol. 31 (7-8) , 285-302
From Promise to Practice: Recent Growth in 30 Years of Tissue Engineering Commercialization
This perspective, marking the 30th anniversary of the Tissue Engineering journal, discusses the exciting trends in the global commercialization of tissue engineering technology. Within a historical context, we present an evolution of challenges and a discussion of the last 5 years of global commercial successes and emerging market trends, highlighting the continued expansion of the field in the northeastern United States. This leads to an overview of the last 5 years’ progress in clinical trials for tissue-engineered therapeutics, including an analysis of trends in success and failure. Finally, we provide a broad overview of preclinical research and a perspective on where the state-of-the-art lies on the horizon.
- 19 February 2025
- Published by Springer Nature in Current Neurology and Neuroscience Reports
- Vol. 25 (1) , 1-12
Machine Learning Approaches to Prognostication in Traumatic Brain Injury
Purpose of Review This review investigates the use of machine learning (ML) in prognosticating outcomes for traumatic brain injury (TBI). It underscores the benefits of ML models in processing and integrating complex, multimodal data—including clinical, imaging, and physiological inputs—to identify intricate non-linear relationships that traditional methods might overlook. Recent Findings ML algorithms of clinical features, neuroimaging, and metrics from the autonomic nervous system enhance the early detection of clinical deterioration and improve outcome prediction. Challenges persist, including issues of data variability, model interpretability, and overfitting. However, advancements in model standardization and validation are key to enhancing their clinical applicability. Summary ML-based, multimodal approaches offer transformative potential for personalized treatment planning and patient management. Future directions include integrating digital twins and real-time continuous data analysis, reinforcing the idea that comprehensive data amalgamation is essential for precise, adaptive prognostication and decision-making in neurocritical care, ultimately leading to better patient outcomes.
- 19 February 2025
- Published by Springer Nature in Current Neurology and Neuroscience Reports
- Vol. 25 (1) , 1-12
Machine Learning Approaches to Prognostication in Traumatic Brain Injury
Purpose of Review This review investigates the use of machine learning (ML) in prognosticating outcomes for traumatic brain injury (TBI). It underscores the benefits of ML models in processing and integrating complex, multimodal data—including clinical, imaging, and physiological inputs—to identify intricate non-linear relationships that traditional methods might overlook. Recent Findings ML algorithms of clinical features, neuroimaging, and metrics from the autonomic nervous system enhance the early detection of clinical deterioration and improve outcome prediction. Challenges persist, including issues of data variability, model interpretability, and overfitting. However, advancements in model standardization and validation are key to enhancing their clinical applicability. Summary ML-based, multimodal approaches offer transformative potential for personalized treatment planning and patient management. Future directions include integrating digital twins and real-time continuous data analysis, reinforcing the idea that comprehensive data amalgamation is essential for precise, adaptive prognostication and decision-making in neurocritical care, ultimately leading to better patient outcomes.
- 5 December 2024
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Geoscience and Remote Sensing
- Vol. 63 (01962892) , 1-27
Hierarchical One-Class Detection for Hyperspectral Image Classification With Background
Hyperspectral image classification (HSIC) has received considerable interest in recent years where most techniques are developed to classify images with background (BKG) removed by ground truth (GT). Unfortunately, in real scenarios, obtaining complete BKG knowledge is generally infeasible. Accordingly, HSIC performed with no BKG (HSIC-NB) is not realistic. Most importantly, many techniques claim to work well for HSIC-NB but perform poorly with BKG included. This article investigates issues arising from BKG in HSIC and further presents a new approach to HSIC with BKG (HSIC-B), called one class detection (OCD), which is based on the well-known hyperspectral subpixel detection technique, constrained energy minimization (CEM). In order for OCD to perform multiclass classification, OCD is further extended to hierarchical OCD (HOC) which is particularly designed to classify multiple classes in a hierarchical tree where each layer uses an iterative kernel CEM (IKCEM) or an iterative kernel target-constrained interference-minimized filter (IKTCIMF) to detect one class at a time for classification. Since M classes are classified by OCD in ${M} -1$ layers in a hierarchical tree, a new concept of class classification priority (CCP) derived from CEM is specifically designed to rank all the classes along the tree in a prioritized order according to their CCP scores. The experimental results demonstrate that hierarchical OCD (HOCD) works well and performs significantly better than many existing HSIC-NB methods at the expense of slightly reduced classification accuracy compared to HSIC-N methods.
- 5 December 2024
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Geoscience and Remote Sensing
- Vol. 63 (01962892) , 1-27
Hierarchical One-Class Detection for Hyperspectral Image Classification With Background
Hyperspectral image classification (HSIC) has received considerable interest in recent years where most techniques are developed to classify images with background (BKG) removed by ground truth (GT). Unfortunately, in real scenarios, obtaining complete BKG knowledge is generally infeasible. Accordingly, HSIC performed with no BKG (HSIC-NB) is not realistic. Most importantly, many techniques claim to work well for HSIC-NB but perform poorly with BKG included. This article investigates issues arising from BKG in HSIC and further presents a new approach to HSIC with BKG (HSIC-B), called one class detection (OCD), which is based on the well-known hyperspectral subpixel detection technique, constrained energy minimization (CEM). In order for OCD to perform multiclass classification, OCD is further extended to hierarchical OCD (HOC) which is particularly designed to classify multiple classes in a hierarchical tree where each layer uses an iterative kernel CEM (IKCEM) or an iterative kernel target-constrained interference-minimized filter (IKTCIMF) to detect one class at a time for classification. Since M classes are classified by OCD in ${M} -1$ layers in a hierarchical tree, a new concept of class classification priority (CCP) derived from CEM is specifically designed to rank all the classes along the tree in a prioritized order according to their CCP scores. The experimental results demonstrate that hierarchical OCD (HOCD) works well and performs significantly better than many existing HSIC-NB methods at the expense of slightly reduced classification accuracy compared to HSIC-N methods.
- Open Access
- 21 October 2024
- Published by Wolters Kluwer Health in Shock
- Vol. 63 (1) , 64-71
COMPARISON OF CONTINUOUS VITAL SIGNS DATA ANALYSIS VERSUS VENOUS LACTATE FOR THE PREDICTION OF LIFESAVING INTERVENTIONS IN PATIENTS WITH TRAUMATIC SHOCK
Introduction: The prehospital environment is fraught with operational constraints, making it difficult to assess the need for resources such as lifesaving interventions (LSI) for adults with traumatic injuries. While invasive methods such as lactate have been found to be highly predictive for estimating injury severity and resource requirements, noninvasive methods, to include continuous vital signs (VS), have the potential to provide prognostic information that can be quickly acquired, interpreted, and incorporated into decision making. In this work, we hypothesized that an analysis of continuous VS would have predictive capacity comparable to lactate and other laboratory tests for the prediction of injury severity, need for LSIs and intensive care unit admission. Methods: In this preplanned secondary analysis of 300 prospectively enrolled patients, venous blood samples were collected in the prehospital environment aboard a helicopter and analyzed with a portable lab device. Patients were transported to the primary adult resource center for trauma in the state of Maryland. Continuous VS were simultaneously collected. Descriptive statistics were used to describe the cohort and predictive models were constructed using a regularized gradient boosting framework with 10-fold cross-validation with additional analyses using Shapley additive explanations (SHAP). Results: Complete VS and laboratory data from 166 patients were available for analysis. The continuous VS models had better performance for prediction of receiving LSIs and intensive care unit length of stay compared to single (initial) VS measurements. The continuous VS models had comparable performance to models using only laboratory tests in predicting discharge within 24 h (continuous VS model: area under the receiver operating curve [AUROC] 0.71; 95% CI, 0.68–0.75 vs. lactate model: AUROC, 0.65; 95% CI, 0.68; 95% CI, 0.66–0.71). The model using all laboratory data yielded the highest sensitivity and sensitivity (AUROC, 0.77; 95% CI, 0.74–0.81). Discussion: The results from this study suggest that continuous VS obtained from autonomous monitors in an aeromedical environment may be helpful for predicting LSIs and the critical care requirements for traumatically injured adults. The collection and use of noninvasively obtained physiological data during the early stages of prehospital care may be useful for in developing user-friendly early warning systems for identifying potentially unstable trauma patients.
- 21 October 2024
- Published by Wolters Kluwer Health in Shock
- Vol. 63 (1) , 64-71
Open AccessCOMPARISON OF CONTINUOUS VITAL SIGNS DATA ANALYSIS VERSUS VENOUS LACTATE FOR THE PREDICTION OF LIFESAVING INTERVENTIONS IN PATIENTS WITH TRAUMATIC SHOCK
Introduction: The prehospital environment is fraught with operational constraints, making it difficult to assess the need for resources such as lifesaving interventions (LSI) for adults with traumatic injuries. While invasive methods such as lactate have been found to be highly predictive for estimating injury severity and resource requirements, noninvasive methods, to include continuous vital signs (VS), have the potential to provide prognostic information that can be quickly acquired, interpreted, and incorporated into decision making. In this work, we hypothesized that an analysis of continuous VS would have predictive capacity comparable to lactate and other laboratory tests for the prediction of injury severity, need for LSIs and intensive care unit admission. Methods: In this preplanned secondary analysis of 300 prospectively enrolled patients, venous blood samples were collected in the prehospital environment aboard a helicopter and analyzed with a portable lab device. Patients were transported to the primary adult resource center for trauma in the state of Maryland. Continuous VS were simultaneously collected. Descriptive statistics were used to describe the cohort and predictive models were constructed using a regularized gradient boosting framework with 10-fold cross-validation with additional analyses using Shapley additive explanations (SHAP). Results: Complete VS and laboratory data from 166 patients were available for analysis. The continuous VS models had better performance for prediction of receiving LSIs and intensive care unit length of stay compared to single (initial) VS measurements. The continuous VS models had comparable performance to models using only laboratory tests in predicting discharge within 24 h (continuous VS model: area under the receiver operating curve [AUROC] 0.71; 95% CI, 0.68–0.75 vs. lactate model: AUROC, 0.65; 95% CI, 0.68; 95% CI, 0.66–0.71). The model using all laboratory data yielded the highest sensitivity and sensitivity (AUROC, 0.77; 95% CI, 0.74–0.81). Discussion: The results from this study suggest that continuous VS obtained from autonomous monitors in an aeromedical environment may be helpful for predicting LSIs and the critical care requirements for traumatically injured adults. The collection and use of noninvasively obtained physiological data during the early stages of prehospital care may be useful for in developing user-friendly early warning systems for identifying potentially unstable trauma patients.
- 19 February 2024
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Geoscience and Remote Sensing
- Vol. 62 (01962892) , 1-22
Iterative Gaussian–Laplacian Pyramid Network for Hyperspectral Image Classification
Gaussian pyramid (GP) is a commonly used image coding technique which encodes an image as a pyramid which is stacked by a set of images with Gaussian window-reduced sizes and multiple spatial resolutions. Associated with GP a Laplacian pyramid (LP) can be also constructed to represent differential images between images in two consecutive layers of GP. Such resulting Gaussian-Laplacian pyramid (GLP) performs data compression in a lossless and lossy manner. A convolutional neural network (CNN) consists of a series of layers concatenated in a feedforward manner where each layer has a convolutional sublayer (CL) and a pooling sublayer (PL). Interestingly, each layer implemented by CL and PL in a CNN can be realized by a single layer in GP in the sense that CL and PL can be carried out by a low-pass Gaussian filter operated as a Gaussian kernel in a single layer of GP. This paper develops a new approach to hyperspectral image classification (HSIC), called Gaussian-Laplacian pyramid network (GLPN) which uses not only GP to realize CNN, but also LP to capture differential information between two consecutive layers that CNN cannot. Furthermore, by incorporating an iterative process into GLPN we can derive an iterative GLPN (IGLPN) that can be considered as a companion of a recently developed iterative random training sampling CNN (IRTS-CNN) by replacing CNN with GLPN. Since GLPN can realize CNN in a better way, it is expected that IGLPN will perform better than IRTS-CNN and also significantly reduce computational efficiency compared to IRTS-CNN.
- 19 February 2024
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Geoscience and Remote Sensing
- Vol. 62 (01962892) , 1-22
Iterative Gaussian–Laplacian Pyramid Network for Hyperspectral Image Classification
Gaussian pyramid (GP) is a commonly used image coding technique which encodes an image as a pyramid which is stacked by a set of images with Gaussian window-reduced sizes and multiple spatial resolutions. Associated with GP a Laplacian pyramid (LP) can be also constructed to represent differential images between images in two consecutive layers of GP. Such resulting Gaussian-Laplacian pyramid (GLP) performs data compression in a lossless and lossy manner. A convolutional neural network (CNN) consists of a series of layers concatenated in a feedforward manner where each layer has a convolutional sublayer (CL) and a pooling sublayer (PL). Interestingly, each layer implemented by CL and PL in a CNN can be realized by a single layer in GP in the sense that CL and PL can be carried out by a low-pass Gaussian filter operated as a Gaussian kernel in a single layer of GP. This paper develops a new approach to hyperspectral image classification (HSIC), called Gaussian-Laplacian pyramid network (GLPN) which uses not only GP to realize CNN, but also LP to capture differential information between two consecutive layers that CNN cannot. Furthermore, by incorporating an iterative process into GLPN we can derive an iterative GLPN (IGLPN) that can be considered as a companion of a recently developed iterative random training sampling CNN (IRTS-CNN) by replacing CNN with GLPN. Since GLPN can realize CNN in a better way, it is expected that IGLPN will perform better than IRTS-CNN and also significantly reduce computational efficiency compared to IRTS-CNN.
- Open Access
- 5 January 2024
- Published by JMIR Publications Inc. in JMIR Human Factors
- Vol. 11 (1) , e46030
A Novel Continuous Real-Time Vital Signs Viewer for Intensive Care Units: Design and Evaluation Study
Background: Clinicians working in intensive care units (ICUs) are immersed in a cacophony of alarms and a relentless onslaught of data. Within this frenetic environment, clinicians make high-stakes decisions using many data sources and are often oversaturated with information of varying quality. Traditional bedside monitors only depict static vital signs data, and these data are not easily viewable remotely. Clinicians must rely on separate nursing charts—handwritten or electric—to review physiological patterns, including signs of potential clinical deterioration. An automated physiological data viewer has been developed to provide at-a-glance summaries and to assist with prioritizing care for multiple patients who are critically ill. Objective: This study aims to evaluate a novel vital signs viewer system in a level 1 trauma center by subjectively assessing the viewer’s utility in a high-volume ICU setting. Methods: ICU attendings were surveyed during morning rounds. Physicians were asked to conduct rounds normally, using data reported from nurse charts and briefs from fellows to inform their clinical decisions. After the physician finished their assessment and plan for the patient, they were asked to complete a questionnaire. Following completion of the questionnaire, the viewer was presented to ICU physicians on a tablet personal computer that displayed the patient’s physiologic data (ie, shock index, blood pressure, heart rate, temperature, respiratory rate, and pulse oximetry), summarized for up to 72 hours. After examining the viewer, ICU physicians completed a postview questionnaire. In both questionnaires, the physicians were asked questions regarding the patient’s stability, status, and need for a higher or lower level of care. A hierarchical clustering analysis was used to group participating ICU physicians and assess their general reception of the viewer. Results: A total of 908 anonymous surveys were collected from 28 ICU physicians from February 2015 to June 2017. Regarding physicians’ perception of whether the viewer enhanced the ability to assess multiple patients in the ICU, 5% (45/908) strongly agreed, 56.6% (514/908) agreed, 35.3% (321/908) were neutral, 2.9% (26/908) disagreed, and 0.2% (2/908) strongly disagreed. Conclusions: Morning rounds in a trauma center ICU are conducted in a busy environment with many data sources. This study demonstrates that organized physiologic data and visual assessment can improve situation awareness, assist clinicians with recognizing changes in patient status, and prioritize care.
- 5 January 2024
- Published by JMIR Publications Inc. in JMIR Human Factors
- Vol. 11 (1) , e46030
Open AccessA Novel Continuous Real-Time Vital Signs Viewer for Intensive Care Units: Design and Evaluation Study
Background: Clinicians working in intensive care units (ICUs) are immersed in a cacophony of alarms and a relentless onslaught of data. Within this frenetic environment, clinicians make high-stakes decisions using many data sources and are often oversaturated with information of varying quality. Traditional bedside monitors only depict static vital signs data, and these data are not easily viewable remotely. Clinicians must rely on separate nursing charts—handwritten or electric—to review physiological patterns, including signs of potential clinical deterioration. An automated physiological data viewer has been developed to provide at-a-glance summaries and to assist with prioritizing care for multiple patients who are critically ill. Objective: This study aims to evaluate a novel vital signs viewer system in a level 1 trauma center by subjectively assessing the viewer’s utility in a high-volume ICU setting. Methods: ICU attendings were surveyed during morning rounds. Physicians were asked to conduct rounds normally, using data reported from nurse charts and briefs from fellows to inform their clinical decisions. After the physician finished their assessment and plan for the patient, they were asked to complete a questionnaire. Following completion of the questionnaire, the viewer was presented to ICU physicians on a tablet personal computer that displayed the patient’s physiologic data (ie, shock index, blood pressure, heart rate, temperature, respiratory rate, and pulse oximetry), summarized for up to 72 hours. After examining the viewer, ICU physicians completed a postview questionnaire. In both questionnaires, the physicians were asked questions regarding the patient’s stability, status, and need for a higher or lower level of care. A hierarchical clustering analysis was used to group participating ICU physicians and assess their general reception of the viewer. Results: A total of 908 anonymous surveys were collected from 28 ICU physicians from February 2015 to June 2017. Regarding physicians’ perception of whether the viewer enhanced the ability to assess multiple patients in the ICU, 5% (45/908) strongly agreed, 56.6% (514/908) agreed, 35.3% (321/908) were neutral, 2.9% (26/908) disagreed, and 0.2% (2/908) strongly disagreed. Conclusions: Morning rounds in a trauma center ICU are conducted in a busy environment with many data sources. This study demonstrates that organized physiologic data and visual assessment can improve situation awareness, assist clinicians with recognizing changes in patient status, and prioritize care.
- 25 December 2023
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Geoscience and Remote Sensing
- Vol. 62 (01962892) , 1-29
Band Sampling of Hyperspectral Anomaly Detection in Effective Anomaly Space
This paper investigates four issues, BKG suppression (BS), anomaly detectability, noise effect, and inter-band correlation reduction (IBCR), which have significant impacts on its performance. Despite that a recently developed effective anomaly space (EAS) was designed to use data sphering (DS) to remove the 2nd order-data statistics characterized BKG, enhance anomaly detectability and reduce noise effect, it does not address the IBCR issue. To cope with this issue, this paper introduces band sampling (BSam) into EAS to reduce IBCR as well as further suppress BKG more effectively. By implementing EAS in conjunction with BSam (EAS-BSam) these four issues can be resolved all together for any arbitrary anomaly detector. It first modifies iterative spectral-spatial hyperspectral anomaly detection (ISSHAD) to develop a new variant of ISSHAD, called iterative spectral-spatial maximal map (ISSMax) and then generalizes ISSMax to EAS-ISSMax which further enhances anomaly detectability and noise removal. Finally, EAS-BSam is implemented to reduce IBCR. As a result, combining EAS, BSam and ISSMax yields four versions, EAS-BSam, EAS-ISSMax, BSam-SSMAX and EAS-BSam-SSMax. Such integration presents a great challenge because all these four versions are derived from different aspects, iterative spectral-spatial feedback process, compressive sensing and low rank and sparse matrix decomposition. Experiments demonstrate that EAS-BSam and EAS-BSam-SSMax show its superiority to ISSHAD and many current existing hyperspectral anomaly detection (HAD) methods.
- 25 December 2023
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Geoscience and Remote Sensing
- Vol. 62 (01962892) , 1-29
Band Sampling of Hyperspectral Anomaly Detection in Effective Anomaly Space
This paper investigates four issues, BKG suppression (BS), anomaly detectability, noise effect, and inter-band correlation reduction (IBCR), which have significant impacts on its performance. Despite that a recently developed effective anomaly space (EAS) was designed to use data sphering (DS) to remove the 2nd order-data statistics characterized BKG, enhance anomaly detectability and reduce noise effect, it does not address the IBCR issue. To cope with this issue, this paper introduces band sampling (BSam) into EAS to reduce IBCR as well as further suppress BKG more effectively. By implementing EAS in conjunction with BSam (EAS-BSam) these four issues can be resolved all together for any arbitrary anomaly detector. It first modifies iterative spectral-spatial hyperspectral anomaly detection (ISSHAD) to develop a new variant of ISSHAD, called iterative spectral-spatial maximal map (ISSMax) and then generalizes ISSMax to EAS-ISSMax which further enhances anomaly detectability and noise removal. Finally, EAS-BSam is implemented to reduce IBCR. As a result, combining EAS, BSam and ISSMax yields four versions, EAS-BSam, EAS-ISSMax, BSam-SSMAX and EAS-BSam-SSMax. Such integration presents a great challenge because all these four versions are derived from different aspects, iterative spectral-spatial feedback process, compressive sensing and low rank and sparse matrix decomposition. Experiments demonstrate that EAS-BSam and EAS-BSam-SSMax show its superiority to ISSHAD and many current existing hyperspectral anomaly detection (HAD) methods.
- 19 July 2023
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Geoscience and Remote Sensing
- Vol. 61 (01962892) , 1-18
Unsupervised Rate Distortion Function-Based Band Subset Selection for Hyperspectral Image Classification
Due to significant inter-band correlation resulting from use of hundreds of contiguous spectral bands, band selection (BS) is one of most widely used methods to reduce data dimensionality for band redundancy removal. A challenge for BS is how to design an effective criterion which can select bands with preserving crucial spectral information, while also avoiding selecting highly correlated bands. Information theory turns out to be one of best means to address such issue in terms of information redundancy, specifically, the rate distortion function (RDF) of Shannon’s 3 rd noisy source coding (or joint source and channel coding) theorem, which has been widely used in image compression/coding. This paper presents a novel unsupervised RDF-based band subset selection (RDFBSS) for hyperspectral image classification (HSIC). To accomplish this goal, a new concept of the area under an RDF curve, A RDF similar to the area under a receiver operating characteristic (ROC), A z defined in hyperspectral target detection is coined and defined as a criterion for BSS. Since BSS generally requires an exhaustive search for an optimal band subset, two iterative algorithms similar to sequential (SQ) N-FINDR and successive (SC) N-FINDR for finding endmembers, called sequential (SQ) RDFBSS and successive (SC) RDFBSS, can be derived and coupled with Ardf as a criterion to find optimal band subsets. The experimental results demonstrate that RDFBSS is indeed a very effective BS method to find best possible band subsets and also performs better than most recent BS methods.
- 19 July 2023
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Geoscience and Remote Sensing
- Vol. 61 (01962892) , 1-18
Unsupervised Rate Distortion Function-Based Band Subset Selection for Hyperspectral Image Classification
Due to significant inter-band correlation resulting from use of hundreds of contiguous spectral bands, band selection (BS) is one of most widely used methods to reduce data dimensionality for band redundancy removal. A challenge for BS is how to design an effective criterion which can select bands with preserving crucial spectral information, while also avoiding selecting highly correlated bands. Information theory turns out to be one of best means to address such issue in terms of information redundancy, specifically, the rate distortion function (RDF) of Shannon’s 3 rd noisy source coding (or joint source and channel coding) theorem, which has been widely used in image compression/coding. This paper presents a novel unsupervised RDF-based band subset selection (RDFBSS) for hyperspectral image classification (HSIC). To accomplish this goal, a new concept of the area under an RDF curve, A RDF similar to the area under a receiver operating characteristic (ROC), A z defined in hyperspectral target detection is coined and defined as a criterion for BSS. Since BSS generally requires an exhaustive search for an optimal band subset, two iterative algorithms similar to sequential (SQ) N-FINDR and successive (SC) N-FINDR for finding endmembers, called sequential (SQ) RDFBSS and successive (SC) RDFBSS, can be derived and coupled with Ardf as a criterion to find optimal band subsets. The experimental results demonstrate that RDFBSS is indeed a very effective BS method to find best possible band subsets and also performs better than most recent BS methods.