A5064837055- Cardiovascular Function and Risk Factors
- Machine Learning in Healthcare
- ECG Monitoring and Analysis
- Atrial Fibrillation Management and Outcomes
- Advanced Neuroimaging Techniques and Applications
- Cardiac Valve Diseases and Treatments
- Cardiovascular Health and Risk Factors
- Artificial Intelligence in Healthcare
- Cardiac Imaging and Diagnostics
- Functional Brain Connectivity Studies
- Genetic Associations and Epidemiology
- Cardiac Arrhythmias and Treatments
- Blind Source Separation Techniques
- Advanced MRI Techniques and Applications
- Phonocardiography and Auscultation Techniques
- Gene expression and cancer classification
- Genomic variations and chromosomal abnormalities
- Spectroscopy and Chemometric Analyses
- Topic Modeling
- Cardiovascular Disease and Adiposity
- Radiomics and Machine Learning in Medical Imaging
- Fractal and DNA sequence analysis
- Cardiac Arrest and Resuscitation
- Bayesian Methods and Mixture Models
- Cardiac pacing and defibrillation studies
Pontifical Catholic University of Peru
2023
Geisinger Medical Center
2018-2022
Geisinger Health System
2018-2021
Mind Research Network
2014-2019
Geisinger Neuroscience Institute
2019
University of New Mexico
2012-2015
Georgia State University
2015
Lovelace Respiratory Research Institute
2014
Deep learning is rapidly advancing many areas of science and technology with multiple success stories in image, text, voice video recognition, robotics, autonomous driving. In this paper we demonstrate how deep neural networks (DNN) trained on large transcriptional response data sets can classify various drugs to therapeutic categories solely based their profiles. We used the perturbation samples 678 across A549, MCF-7, PC-3 cell lines from LINCS Project linked those 12 use derived MeSH. To...
Background: Atrial fibrillation (AF) is associated with substantial morbidity, especially when it goes undetected. If new-onset AF could be predicted, targeted screening used to find early. We hypothesized that a deep neural network predict from the resting 12-lead ECG and this prediction may help identify those at risk of AF-related stroke. Methods: 1.6 M digital traces 430 000 patients collected 1984 2019. Deep networks were trained (within 1 year) in without history AF. Performance was...
Timely diagnosis of structural heart disease improves patient outcomes, yet many remain underdiagnosed. While population screening with echocardiography is impractical, ECG-based prediction models can help target high-risk patients. We developed a novel machine learning approach to predict multiple conditions, hypothesizing that composite model would yield higher prevalence and positive predictive values facilitate meaningful recommendations for echocardiography.Using 2 232 130 ECGs linked...
The development of objective biomarkers for mild traumatic brain injury (mTBI) in the chronic period is an important clinical and research goal. Head trauma known to affect mechanisms that support electrophysiological processing information within between regions, so methods like quantitative EEG may provide viable indices dysfunction associated with even mTBI.Resting-state, eyes-closed data were obtained from 71 individuals military-related mTBI 82 normal comparison subjects without injury....
The majority of biomedical studies use limited datasets that may not generalize over large heterogeneous have been collected several decades. current paper develops and validates multimodal models can predict 1-year mortality based on a massive clinical dataset. Our focus predicting provide sense urgency to the patients. Using largest dataset its kind, considers development validation 25,137,015 videos associated with 699,822 echocardiography from 316,125 patients, 2,922,990 8-lead...
Despite the rapidly growing interest, progress in study of relations between physiological abnormalities and mental disorders is hampered by complexity human brain high costs data collection. The can be captured deep learning approaches, but they still may require significant amounts data. In this paper, we seek to mitigate latter challenge developing a generator for synthetic realistic training Our method greatly improves generalization classification schizophrenia patients healthy controls...
Deep learning methods have significantly improved classification accuracy in different areas such as speech, object and text recognition. However, this field has only began to be explored the brain imaging field, which differs from other fields terms of amount data available, its dimensionality factors. This paper proposes a methodology generate an extensive synthetic structural magnetic resonance (sMRI) dataset used at pre-training stage shallow network model address issue having limited...
Deep learning algorithms, in particular long short-term memory (LSTM), have become an increasingly popular choice for natural language processing a variety of applications such as sentiment analysis and text analysis. In this study, we propose fully automated deep algorithm which learns to classify radiological reports the presence intracranial hemorrhage (ICH) diagnosis. The proposed architecture consists 1D convolution neural networks (CNN), (LSTM) units logistic function was trained...
The association of copy number variation (CNV) with schizophrenia has been reported evidence increased frequency both rare and large CNVs. Yet, little is known about the impact CNVs in brain structure. In this pilot study, we explored collective effects all each cytogenetic band on risk gray matter measured structural magnetic resonance imaging. With 324 participants’ CNV profiles (151 patients 173 healthy controls), first extracted specific features that differ between controls using a two...
The wide variety of brain imaging technologies allows us to exploit information inherent different data modalities. richness multimodal datasets may increase predictive power and reveal latent variables that otherwise would have not been found. However, the analysis is often conducted by assuming linear interactions which impact accuracy results. We propose use a multi-layer perceptron model enhance structural functional magnetic resonance (sMRI fMRI) combined. also synthetic generator...
Background Copy number variations ( CNV s) are structural genetic mutations consisting of segmental gains or losses in DNA sequence. Although s contribute substantially to genomic variation, few and imaging studies report association with alcohol dependence AD ). Our purpose is find evidence this across ethnic populations genders. This work the first – study groups include A frican merican (AA) population. Methods considers 2 data sets, one for discovery (2,345 samples) other validation (239...
High data dimensionality poses a major challenge for imaging genomic studies. To address this issue, semi-blind multivariate approach, parallel independent component analysis with multiple references (pICA-MR), is proposed. pICA-MR extracts and genetic components in enhances inter-modality correlations. Prior knowledge incorporated to emphasize factors specific attributes. Particularly, can investigate explore functional interactions among genes. Simulations demonstrate robust performances...
We present an interpretable neural network for predicting important clinical outcome (1-year mortality) from multi-modal Electronic Health Record (EHR) data. Our approach builds on prior machine learning models by now enabling visualization of how individual factors contribute to the overall risk, assuming other remain constant, which was previously impossible. demonstrate value this using a large dataset including both EHR data and 31,278 echocardiographic videos heart 26,793 patients....
Several large trials have employed age or clinical features to select patients for atrial fibrillation (AF) screening reduce strokes. We hypothesized that a machine learning (ML) model trained predict AF risk from 12‑lead electrocardiogram (ECG) would be more efficient than criteria based on variables in indicating population potentially prevent AF-related stroke.We retrospectively included all with encounters Geisinger without prior history of AF. Incidence within 1 year and strokes 3 years...
Abstract Imputation is a key step in Electronic Health Records-mining as it can significantly affect the conclusions derived from downstream analysis. There are three main categories that explain missingness clinical settings–incompleteness, inconsistency, and inaccuracy–and these capture variety of situations: patient did not seek treatment, health care provider enter information, etc. We used EHR data patients diagnosed with Inflammatory Bowel Disease Geisinger System to design novel...
Electronic health records (EHR) contain a large variety of information on the clinical history patients such as vital signs, demographics, diagnostic codes and imaging data. The enormous potential for discovery in this rich dataset is hampered by its complexity heterogeneity. We present first study to assess unsupervised homogenization pipelines designed EHR clustering. To identify optimal pipeline, we tested accuracy simulated data with varying amounts redundancy, heterogeneity,...