Welcome
Azka Rehman
AI Researcher & Medical Imaging Specialist
Developing deep learning systems for opportunistic screening in medical imaging. From detecting cardiovascular risk in routine chest X-rays to finding lung nodules in CT scans. 5+ publications, 5+ imaging modalities, bridging AI and clinical practice.
About Me
I am an AI researcher focused on developing deep learning systems for medical image analysis. My work centers on opportunistic screening: extracting clinically meaningful biomarkers from routine imaging studies such as chest X-rays and abdominal CT scans to identify early signs of cardiovascular disease.
My research is at the intersection of computer vision and clinical medicine, with a focus on foundation models, explainability, and survival analysis for risk stratification. The goal is to make every routine scan count, especially in settings where specialized imaging is not available.
I completed my M.S. at Seoul National University and spent 2+ years at HealthHub.kr deploying clinical AI products used by radiologists in practice. I have worked across 5+ imaging modalities (CT, MRI, X-ray, ultrasound, retinal fundus) and published in journals including npj Digital Medicine, IEEE JBHI, and CMIG.
Education & Experience
Education
M.S. in Biomedical Sciences
Seoul National University (SNU)
2023 - 2025 | Seoul, South Korea
Thesis: AI-Driven Opportunistic Screening for Cardiovascular Disease Risk Using Abdominal CT Scans
- ▷Health System Data Science Lab, supervised by Dr. Sang Min Park.
- ▷Research on foundation models for opportunistic cardiovascular screening from medical images.
- ▷Published in CMIG, npj Digital Medicine, and IEEE JBHI.
B.S. in Electrical Engineering
National University of Science and Technology (NUST)
2016 - 2020 | Islamabad, Pakistan
Thesis: Chest X-Ray Abnormality Detection using Deep Learning
- ▷Focus on signal processing and machine learning.
- ▷Published brain tumor segmentation work (Sensors 2023).
Work Experience
Researcher
Health System Data Science Lab, Seoul National University
Sep 2023 - Aug 2025 | Seoul, South Korea
- ▷Developed opportunistic screening pipelines for cardiovascular risk prediction from chest X-rays and abdominal CT scans.
- ▷Evaluated foundation models (Rad-DINO, DINOv2, CheXagent, OpenCLIP) with LoRA fine-tuning for atherosclerosis detection.
- ▷Conducted survival analysis linking imaging biomarkers to cardiovascular mortality outcomes.
AI Research Engineer
HealthHub.kr
Jan 2021 - Jul 2023 | Seoul, South Korea
- ▷Developed deep learning solutions for lung nodule detection and medical image analysis deployed on the DICOMLINK/HPACS platform.
- ▷Integrated AI pipelines with clinical systems, optimizing for efficiency and scalability.
- ▷Mentored and trained new engineers on ML/DL workflows for medical imaging.
Publications
Opportunistic Screening of Carotid Atherosclerosis and Cardiovascular Mortality Risk Using Chest X-Rays: A Comparative Study of Foundation Models
A. Rehman, J. Kim, H. Lee, J. Chang, S. Park
Journal of American Heart Association
SSMT-Net: A Semi-Supervised Multitask Transformer-Based Network for Thyroid Nodule Segmentation
M. U. Farooq, A. Ur Rehman, A. Rehman, M. Usman, D. K. Chae
WACV 2026
Projects
AI-Driven Cardiovascular Risk Stratification from Abdominal CT
Developed an opportunistic screening pipeline to estimate cardiovascular risk from routine abdominal CT scans. Deep learning models for imaging-based risk prediction evaluated using survival analysis.
Atherosclerosis Screening from Chest X-Rays with Foundation Models
Comparative evaluation of foundation models (Rad-DINO, DINOv2, CheXagent, CLIP) for opportunistic cardiovascular risk prediction from standard chest X-rays with explainability analysis.
Open Source & GitHub
MESAHA-Net
Medical image segmentation and analysis network.
SDS-MSA-Net
Fully automatic method for segmenting brain tumor regions using attention network.
KL Grade Classification in Knee X-ray Images
Novel architectures for Kellgren-Lawrence grade classification from knee X-rays.
MAXedNet
Deep learning network architecture for medical image analysis.