Healthcare
Earth Observation
Industry 5.0
Diagnostics
Clinical Trials
Cardiac Imaging
Advancing Early Heart Attack Detection through Plaque Activity Analysis
December 23, 2025
Lightsource Research
Lightsource Research is a cardiac imaging research company advancing AI-assisted analysis of coronary CT angiograms (CCTA) to improve cardiovascular risk assessment.
Undetected active plaques leads to 7 times higher risk of heart attack
The difficulty of accurately quantifying and characterizing "vulnerable" or "high-risk" plaques (HRP) that are prone to rupture but often missed by traditional stenosis (blockage) grading.
Project Goals
- Enable AI models to generate highly accurate virtual maps of coronary artery plaque burden
- Reduce time-to-model by streamlining the manual assessment of large volumes of CCTA scans
- Deliver clinically reliable training datasets to support scalable, AI-driven cardiac risk prediction
Introduction
Lightsource Research is a cardiac imaging research company advancing AI-assisted analysis of coronary CT angiograms (CCTA) to improve cardiovascular risk assessment.
Main Challenge
The difficulty of accurately quantifying and characterizing "vulnerable" or "high-risk" plaques (HRP) that are prone to rupture but often missed by traditional stenosis (blockage) grading.
Goals
- Enable AI models to generate highly accurate virtual maps of coronary artery plaque burden
- Reduce time-to-model by streamlining the manual assessment of large volumes of CCTA scans
- Deliver clinically reliable training datasets to support scalable, AI-driven cardiac risk prediction
Identifying early plaque activity to enable strong preventative care
Through radiologist-led annotation workflows, Ingedata delivered plaque-level datasets that support earlier and more informed cardiac risk assessment.
Project Impact
Improved detection of active plaque burden
Enhanced support for preventative treatment planning
Reduced risk of unexpected heart attack events
Creating clinically reliable plaque activity datasets for cardiac AI
Clinical Dataset Selection
Selected 300 coronary CT angiogram scans from clinical trial datasets, ensuring diagnostic relevance, image quality and protocol consistency.
0
1
Clinical Dataset Selection
Coronary Anatomy Segmentation
Created detailed masks of coronary arteries, lumen and vessel walls to accurately represent complex coronary anatomy.
0
2
Coronary Anatomy Segmentation
Plaque Characterisation
Manually segmented calcified, non-calcified and low-attenuation plaques to capture clinically meaningful plaque features.
0
3
Plaque Characterisation
Tissue Differentiation Standards
Applied consistent Hounsfield Unit (HU) windowing to reliably distinguish vessel walls, lumen and plaque tissue.
0
4
Tissue Differentiation Standards
Radiologist Review and Validation
Cardiac radiologists reviewed and corrected annotations to ensure accuracy, consistency and clinical usability.
0
5
Radiologist Review and Validation
No items found.
