Related Experiment Video
Updated: Jun 10, 2025
05:32
Retrospective Cardiac Gating with A Prototype Small-Animal X-ray Computed Tomograph
Published on: February 21, 2025
237
Bolus tracking from pulsed x-ray projections: A feasibility study using a five-dimensional cardiac CT contrast
Eri Haneda1, Isabelle M Heukensfeldt Jansen1, Pengwei Wu1
1GE HealthCare Technology and Innovation Center, Niskayuna, New York, USA.
Medical Physics
|October 16, 2024
Summary
This study introduces a novel method using AI and virtual data to accurately predict contrast peak timing in cardiac CT scans, improving diagnostic efficiency. The technique enhances autonomous cardiac CT workflows by precisely triggering scans for optimal contrast enhancement.
Area of Science:
- Medical Imaging
- Artificial Intelligence
- Cardiovascular Imaging
Background:
- Cardiac CT requires precise timing for optimal contrast enhancement, posing a challenge for current bolus tracking methods.
- Existing techniques for timing contrast bolus in cardiac CT have limitations in reducing contrast volume, scan time, and manual interventions.
- Improving the robustness of peak contrast opacification timing is crucial for complex cardiac CT examinations.
Purpose of the Study:
- To develop an autonomous cardiac CT workflow for tracking contrast dynamics directly from pulsed x-ray projections.
- To create a novel 5D virtual cardiac CT data generation tool for simulating realistic cardiac profiles and bolus dynamics.
- To demonstrate projection-domain prospective bolus tracking using a neural network trained on virtual data for peak contrast identification.
Main Methods:
- Acquired pulsed mode projections (PMPs) under sparse view conditions for real-time contrast enhancement estimation.
- Developed a 5D cardiac model to generate clinically realistic virtual scan data with programmable cardiac and bolus dynamics.
- Trained and evaluated neural networks using virtual data, exploring 20 projection angles and 300 virtual exams per angle.
Main Results:
- Neural network estimation achieved a cosine similarity greater than 0.97 compared to ground truth for bolus time-intensity curves.
- Accurate prediction of the contrast bolus curve shape, with low root-mean-square error (RMSE) for peak time estimation (1.23s left, 0.78s right chambers).
- Identified an optimal projection angle (30 degrees from vertical) for minimizing errors in bolus peak time estimation.
Conclusions:
- An innovative real-time method for predicting contrast bolus peaks in cardiac CT was proposed.
- A novel approach using virtual, clinically realistic data effectively trained a neural network for accurate contrast peak estimation.
- The developed technique shows potential for autonomous cardiac CT, enabling scan triggering for optimal contrast enhancement.
