Researchers have developed a unified framework for cardiac CT segmentation and phenotyping, combining a human-in-the-loop annotation process with a self-supervised foundation model. This approach, pre-trained on 60,000 unlabeled scans, has produced the largest expert-annotated cardiac CT dataset to date, with 1598 cases. The framework demonstrated superior accuracy and efficiency compared to existing tools, particularly in low-data scenarios, and highlighted the importance of data quality and pre-training over specific architectures. AI
IMPACT This framework could accelerate opportunistic cardiac phenotyping from routine CT scans, providing clinically relevant insights into ventricular function and disease severity.
RANK_REASON The cluster contains a research paper detailing a new framework and dataset for medical image analysis.
- arXiv
- Cardiac-CT in the Treatment of Acute Chest Pain
- computed tomography
- convolutional neural network
- Hugging Face
- interactive machine learning
- phase space
- self-supervised foundation model
- state-space architectures
- Transformer++
- vision foundation model
AI-generated summary · Google Gemini · from 2 sources. How we write summaries →
