A5064289969- Trauma and Emergency Care Studies
- Traumatic Brain Injury and Neurovascular Disturbances
- ECG Monitoring and Analysis
- Intracerebral and Subarachnoid Hemorrhage Research
- EEG and Brain-Computer Interfaces
- Chronic Obstructive Pulmonary Disease (COPD) Research
- Acute Ischemic Stroke Management
- Computational Physics and Python Applications
- Sepsis Diagnosis and Treatment
- Autopsy Techniques and Outcomes
- Respiratory Support and Mechanisms
- Medical Image Segmentation Techniques
- Genetics, Bioinformatics, and Biomedical Research
- Medical Imaging and Analysis
- Cardiac electrophysiology and arrhythmias
- Advanced X-ray and CT Imaging
- Radiation Dose and Imaging
- Neurosurgical Procedures and Complications
- Non-Invasive Vital Sign Monitoring
- Emergency and Acute Care Studies
- Occupational and environmental lung diseases
- Medical and Biological Ozone Research
- Machine Learning in Healthcare
- Image and Object Detection Techniques
- Sexual function and dysfunction studies
University of Michigan
2016-2024
Abstract There is a growing gap between studies describing the capabilities of artificial intelligence (AI) diagnostic systems using deep learning versus efforts to investigate how or when integrate AI into real-world clinical practice support physicians and improve diagnosis. To address this gap, we four potential strategies for model deployment physician collaboration determine their impact on accuracy. As case study, examine an trained identify findings acute respiratory distress syndrome...
Abstract Prognosis of the long-term functional outcome traumatic brain injury is essential for personalized management that injury. Nonetheless, accurate prediction remains unavailable. Although machine learning has shown promise in many fields, including medical diagnosis and prognosis, such models are rarely deployed real-world settings due to a lack transparency trustworthiness. To address these drawbacks, we propose learning-based framework explainable aligns with clinical domain...
Objectives/Goals: The objective of this study is to explore strategies for AI-physician collaboration in diagnosing acute respiratory distress syndrome (ARDS) using chest X-rays. By comparing the diagnostic accuracy different AI deployment methods, aims identify optimal that leverage both and physician expertise improve outcomes. Methods/Study Population: analyzed 414 frontal X-rays from 115 patients hospitalized between August 15 October 2, 2017, at University Michigan. Each X-ray was...
Abstract Background Both early detection and severity assessment of liver trauma are critical for optimal triage management patients. Current protocols utilize computed tomography (CT) injuries in a subjective qualitative (v.s. quantitative) fashion, shortcomings which could both be addressed by automated computer-aided systems that capable generating real-time reproducible quantitative information. This study outlines an end-to-end pipeline to calculate the percentage parenchyma disrupted...
Detection and severity assessment of subdural hematoma is a major step in the evaluation traumatic brain injuries. This retrospective study 110 computed tomography (CT) scans from patients admitted to Michigan Medicine Neurological Intensive Care Unit or Emergency Department. A machine learning pipeline was developed segment assess hematoma. First, probability each point belonging region determined using combination hand-crafted deep features. provided initial state segmentation. Next, 3D...
OBJECTIVES: Implementing a predictive analytic model in new clinical environment is fraught with challenges. Dataset shifts such as differences practice, data acquisition devices, or changes the electronic health record (EHR) implementation mean that input seen by can differ significantly from it was trained on. Validating models at multiple institutions therefore critical. Here, using retrospective data, we demonstrate how Predicting Intensive Care Transfers and other UnfoReseen Events...
Traumatic brain injury is a serious public health problem in the U.S. contributing to large portion of permanent disability. However, its early management and treatment could limit impact injury, save lives reduce burden cost for patients as well healthcare systems. Subdural hematoma one most common types TBI, which visual detection quantitative evaluation are time consuming prone error. In this study, we propose fully auto-mated machine learning based approach 3D segmentation convexity...
Traumas and illnesses can cause injury in internal organs. The liver, being the largest abdominal organ, is most likely to be injured by trauma. Currently CT scans are analyzed radiologists see if there any injuries organs; however, due large amounts of data its complexity terms noise, intensity variations different images so on, visual inspection would time consuming prone error. Therefore, an automated approach beneficial. In this paper we propose a fully Bayesian based method for 3D...
Abstract Purpose Pediatric acute respiratory distress syndrome (PARDS) is underrecognized in the pediatric intensive care unit and interpretation of chest radiographs a key step identification. We sought to test performance machine learning model detect PARDS cohort children with failure. Materials methods A convolutional neural network (CNN) previously developed ARDS on adult was applied age 7 days 18 years, admitted PICU, mechanically ventilated through tracheostomy, endotracheal tube or...
Traumatic abdominal injury can lead to multiple complications including laceration of major organs such as kidneys. Contrast-enhanced Computed Tomography (CT) is the primary imaging modality for evaluating kidney injury. However, traditional visual examination CT scans time consuming, non-quantitative, prone human error, and costly. In this work we propose a segmentation method using machine learning active contour modeling. We first detect an initialization mask inside then evolve its...
In a variety of injuries and illnesses, internal organs in the abdominal pelvic regions, particular liver, may be compromised. current practice, CT scans liver are visually inspected to investigate integrity organ. However, size complexity images limits reliability visual inspection accurately assess health liver. Computer-aided image analysis can create fast quantitative assessment from CT, environments where access skilled radiologists limited. this paper we propose hierarchical method...
Traumatic brain injury (TBI) is a major health and socioeconomic problem globally that associated with high level of mortality. Early accurate diagnosis prognosis TBI important in patient management preventing any secondary injuries. Computer tomography (CT) imaging assists physicians diagnosing guiding treatment. One the clinical parameters extracted from CT images midline shift, measure linear displacement structure, which correlated outcomes. However, only tiny fraction overall tissue...
Automated segmentation of the spleen in CT volumes is difficult due to variations size, shape, and position within abdominal cavity as well similarity intensity values among organs cavity. In this paper we present a method for automated localization axial using trained classification models, active contours, anatomical information, adaptive features. The results show an average Dice score 0.873 on patients experiencing various chest, abdominal, pelvic traumas taken at different contrast phases.
We are bioinformatics trainees at the University of Michigan who started a local chapter Girls Who Code to provide fun and supportive environment for high school women learn power coding. Our goal was cover basic coding topics data science concepts through live hands-on practice. However, we could not find resource that exactly met our needs. Therefore, over past three years, have developed curriculum instructional format using Jupyter notebooks effectively teach introductory Python science....
This paper presents a novel sensor in the form of ring to monitor vascular tone continuously and non-invasively. The signal that is generated by this would allow for accurate detection analysis reflection waves can be used identify or predict several medical conditions including intradialytic hypotension (IDH). accompanied analytics detect changes morphology which are associated with IDH notify dialysis physicians nurses. data was collected from nine patients indicated proposed processing...
The QRS complex is the most prominent feature of electrocardiogram (ECG) that used as a marker to identify cardiac cycles. Identification locations enables arrhythmia detection and heart rate variability estimation. Therefore, accurate consistent localization an important component automated ECG analysis which necessary for early cardiovascular diseases. This study evaluates performance six popular publicly available methods on large dataset over half million ECGs in diverse population...
The PR interval represents the time required from electrical impulse to advance atrium AV node and His-Purkinje system until ventricular myocardium begins depolarize.PR prolongation has been associated with significant increases in atrial fibrillation, heart failure mortality.Over past years, multiple deep learning models have proposed interpret electrocardiogram (ECG) signals.Despite initial success, these are often trained validated using datasets that contain partially incorrect...