Synthesizing CTA Image Data for Type-B Aortic Dissection using Stable Diffusion Models.

Area of Science:

• Artificial Intelligence
• Medical Imaging
• Cardiovascular Disease

Background:

• Generative AI, specifically Stable Diffusion (SD), shows promise for synthesizing medical imaging data.
• Data scarcity is a significant limitation in training machine learning (ML) algorithms for cardiovascular image processing.

Purpose of the Study:

• To explore the generation of synthetic cardiac Computed Tomography Angiography (CTA) images using fine-tuned Stable Diffusion models.
• To address the challenge of limited data availability in cardiovascular imaging research.

Main Methods:

• Fine-tuning Stable Diffusion models with a limited dataset of cardiac CTA images.
• Utilizing user-defined text prompts for image generation.
• Conducting comprehensive evaluations including quantitative analysis and qualitative assessment by a clinician.

Main Results:

• Successful generation of synthetic cardiac CTA images using a Text-to-Image (T2I) Stable Diffusion model.
• The optimized T2I CTA diffusion model effectively rendered images with features characteristic of acute type B aortic dissection (TBAD).

Conclusions:

• Text-to-Image Stable Diffusion models are capable of generating realistic cardiac CTA images.
• This approach can help overcome data scarcity issues in cardiovascular imaging, enhancing ML algorithm development for conditions like TBAD.