A5091101940- Machine Learning in Healthcare
- Dementia and Cognitive Impairment Research
- Biomedical Text Mining and Ontologies
- Artificial Intelligence in Healthcare
- Explainable Artificial Intelligence (XAI)
- Functional Brain Connectivity Studies
- COVID-19 diagnosis using AI
- Radiomics and Machine Learning in Medical Imaging
- Statistical Methods in Clinical Trials
- Advanced Neuroimaging Techniques and Applications
- Medical Imaging Techniques and Applications
- Topic Modeling
- Artificial Intelligence in Healthcare and Education
- Neural Networks and Applications
- Mental Health and Psychiatry
- Bioinformatics and Genomic Networks
- Advanced MRI Techniques and Applications
- Organ Transplantation Techniques and Outcomes
- Health Systems, Economic Evaluations, Quality of Life
- Alzheimer's disease research and treatments
- Stress Responses and Cortisol
- Estrogen and related hormone effects
- Transplantation: Methods and Outcomes
- Chronic Disease Management Strategies
- Generative Adversarial Networks and Image Synthesis
Washington University in St. Louis
2021-2025
Mallinckrodt (United States)
2024
Primary graft dysfunction (PGD) is a common complication after lung transplantation associated with poor outcomes. Although risk factors have been identified, the complex interactions between clinical variables affecting PGD are not well understood, which can complicate decisions about donor acceptance. Previously, we developed machine learning (ML) model to predict grade 3 using and recipient electronic health record (EHR) data, but it lacked granular information from CT scans, routinely...
Multiple signal modalities, such as vision and sounds, are naturally present in real-world phenomena. Recently, there has been growing interest learning generative models, particular variational autoencoder (VAE), to for multimodal representation especially the case of missing modalities. The primary goal these models is learn a modality-invariant modality-specific that characterizes information across multiple Previous attempts at VAEs approach this mainly through lens experts, aggregating...
In the past decades, deep neural networks, particularly convolutional have achieved state-of-the-art performance in a variety of medical image segmentation tasks. Recently, introduction vision transformer (ViT) has significantly altered landscape models. There been growing focus on ViTs, driven by their excellent and scalability. However, we argue that current design transformer-based UNet (ViT-UNet) models may not effectively handle heterogeneous appearance (e.g., varying shapes sizes)...
Normative models in neuroimaging learn the brain patterns of healthy population distribution and estimate how disease subjects like Alzheimer's Disease (AD) deviate from norm. Existing variational autoencoder (VAE)-based normative using multimodal data aggregate information multiple modalities by estimating product or averaging unimodal latent posteriors. This can often lead to uninformative joint distributions which affects estimation subject-level deviations. In this work, we addressed...
We aimed to develop and validate a novel multimodal framework
ABSTRACT Differences in amyloid positron emission tomography (PET) radiotracer pharmacokinetics and binding properties lead to discrepancies amyloid‐β uptake estimates. Harmonization of tracer‐specific biases is crucial for optimal performance downstream tasks. Here, we investigated the efficacy ComBat, a data‐driven harmonization model, reducing regional PET measurements from [ 18 F]‐florbetapir (FBP) 11 C]‐Pittsburgh compound‐B (PiB). One hundred thirteen head‐to‐head FBP‐PiB scan pairs,...
Normative modelling is a method for understanding the underlying heterogeneity within brain disorders like Alzheimer Disease (AD), by quantifying how each patient deviates from expected normative pattern that has been learned healthy control distribution. Existing deep learning based models have applied on only single modality Magnetic Resonance Imaging (MRI) neuroimaging data. However, these do not take into account complementary information offered multimodal M RI, which essential...
Dementia is characterized by a decline in memory and thinking that significant enough to impair function activities of daily living. Patients seen dementia specialty clinics are highly heterogenous with variety different symptoms progress at rates. Recent research has focused on finding data-driven subtypes for revealing new insights into dementia’s underlying heterogeneity, rather than assuming the cohort homogenous. However, current studies subtyping have following limitations: (i)...
Abstract The risk of Alzheimer’s disease (AD) in women is about 2 times greater than men. estrogen hypothesis being accepted as the essential sex factor causing difference AD. Also, recent meta-analysis using large-scale medical records data indicated replacement therapy. However, underlying molecular targets and mechanisms explaining this AD development remain unclear. In study, we identified that treatment can strongly inhibition neuro-inflammation signaling targets, systems pharmacology...
Complex deep learning models show high prediction tasks in various clinical but their inherent complexity makes it more challenging to explain model predictions for clinicians and healthcare providers. Existing research on explainability of have two major limitations: using post-hoc explanations raw variables as units explanation, both which are often difficult human interpretation. In this work, we designed a self-explaining framework the expert-knowledge driven concepts or intermediate...
Structured Abstract INTRODUCTION Previous studies have applied normative modeling on a single neuroimaging modality to investigate Alzheimer Disease (AD) heterogeneity. We employed deep learning-based multimodal framework analyze individual-level variation across ATN (amyloid-tau-neurodegeneration) imaging biomarkers. METHODS selected cross-sectional discovery (n = 665) and replication cohorts 430) with available T1-weighted MRI, amyloid tau PET. Normative estimated abnormal deviations in...
Early identification of Mild Cognitive Impairment (MCI) subjects who will eventually progress to Alzheimer Disease (AD) is challenging. Existing deep learning models are mostly single-modality single-task predicting risk disease progression at a fixed timepoint. We proposed multimodal hierarchical multi-task approach which can monitor the each timepoint visit trajectory. Longitudinal data from multiple modalities (MRI, cognition, and clinical data) were collected MCI individuals Neuroimaging...
Abstract Dementia is characterized by a decline in memory and thinking that significant enough to impair function activities of daily living. Patients seen dementia specialty clinics are highly heterogenous with variety different symptoms progress at rates. Recent research has focused on finding data-driven subtypes for revealing new insights into dementia’s underlying heterogeneity, compared analyzing the entire cohort as single homogeneous group. However, current studies subtyping have...
Differences in amyloid positron emission tomography (PET) radiotracer pharmacokinetics and binding properties lead to discrepancies amyloid-β uptake estimates. Harmonization of tracer-specific biases is crucial for optimal performance downstream tasks. Here, we investigated the efficacy ComBat, a data-driven harmonization model, reducing regional PET measurements from [
Abstract Normative models in neuroimaging learn patterns of healthy brain distributions to identify deviations disease subjects, such as those with Alzheimer’s Disease (AD). This study addresses two key limitations variational autoencoder (VAE)-based normative models: (1) VAEs often struggle accurately model control distributions, resulting high reconstruction errors and false positives, (2) traditional multimodal aggregation methods, like Product-of-Experts (PoE) Mixture-of-Experts (MoE),...
Multiple signal modalities, such as vision and sounds, are naturally present in real-world phenomena. Recently, there has been growing interest learning generative models, particular variational autoencoder (VAE), to for multimodal representation especially the case of missing modalities. The primary goal these models is learn a modality-invariant modality-specific that characterizes information across multiple Previous attempts at VAEs approach this mainly through lens experts, aggregating...
Abstract Background Differences in amyloid PET radiotracer pharmacokinetics and binding properties lead to discrepancies uptake measurements, which may adversely affect the statistical power of clinical trials that utilize multiple tracers track brain deposition. To address this, Centiloid was developed for standardizing global SUVRs across a common scale. Alternatively, ComBat is technique harmonizing batch effects while preserving variations from biologically‐relevant covariates. Unlike...
Abstract Background Differences in amyloid PET radiotracer pharmacokinetics and binding properties lead to discrepancies uptake measurements, which may adversely affect the statistical power of clinical trials that utilize multiple tracers track brain deposition. To address this, Centiloid was developed for standardizing global SUVRs across a common scale. Alternatively, ComBat is technique harmonizing batch effects while preserving variations from biologically‐relevant covariates. Unlike...
Normative modelling is an emerging method for understanding the underlying heterogeneity within brain disorders like Alzheimer Disease (AD) by quantifying how each patient deviates from expected normative pattern that has been learned a healthy control distribution. Since AD multifactorial disease with more than one biological pathways, multimodal magnetic resonance imaging (MRI) neuroimaging data can provide complementary information about heterogeneity. However, existing deep learning...
Normative models in neuroimaging learn the brain patterns of healthy population distribution and estimate how disease subjects like Alzheimer's Disease (AD) deviate from norm. Existing variational autoencoder (VAE)-based normative using multimodal data aggregate information multiple modalities by estimating product or averaging unimodal latent posteriors. This can often lead to uninformative joint distributions which affects estimation subject-level deviations. In this work, we addressed...