Medical Imaging AI
Acute Stroke Imaging
Self-supervised and multimodal learning for 3D CTA, NCCT, and clinical outcome prediction in acute stroke.
Yingjun Dong, Ph.D.
Postdoctoral Research Fellow
UTHealth Houston
Model Development for Learning, Memory, and Spatial Navigation
I develop computational and machine learning models to understand how the brain represents space and context, and to advance medical imaging AI for acute stroke diagnosis and outcome prediction.
Research Focus
Featured Publications
View all publications ->Generalizable self-supervised learning for brain CTA in acute stroke
Dong Y, Pachade S, Roberts K, Jiang X, Sheth S A, Giancardo L.
↗Can CTA-Based Machine Learning Identify Patients for Whom Successful Endovascular Stroke Therapy Is Insufficient?
Jeevarajan J A, Dong Y, Ballekere A, Marioni S S, Niktabe A, et al.
↗A self-supervised learning approach for registration agnostic imaging models with 3D brain CTA
Dong Y, Pachade S, Liang X, Sheth S A, Giancardo L.
↗Self-supervised learning with radiology reports, a comparative analysis of strategies for large vessel occlusion and brain CTA images
Pachade S, Datta S, Dong Y, Salazar-Marioni S, Abdelkhaleq R, et al.
↗Selected Projects
Projects selected from my resume, spanning acute stroke imaging AI, image-text learning, and multimodal human behavior analysis.
Generalizable Self-Supervised Learning for Brain CTA
Image-text pretraining with 3D CTA and radiology reports for acute stroke prediction tasks.
Registration-Agnostic Imaging Models with 3D CTA
Self-supervised models designed to reduce dependence on time-consuming registration preprocessing.
Image-Text Paired Medical Data
LLaMA-assisted radiology report summarization and paired imaging-report representation learning.
Utterance Clustering on Real-World Audio Data
Speech/audio preprocessing and Gaussian mixture modeling for utterance clustering.
Feature Selection on Facial Landmarks
Hybrid information-theory, clustering, and genetic algorithm feature selection for facial emotion recognition.
Narrative Background Analysis with Doc2Vec
Text representation analysis for background diversity in collective ideation experiments.
Open to research collaborations in AI, imaging, and neuroscience.
Yingjun.Dong@uth.tmc.edu