A German team has released TAIX-Ray, a massive bedside chest x-ray dataset supporting AI development for pathology detection in ICU settings.
Key Details
- 1TAIX-Ray dataset includes 215,381 chest x-rays from 47,724 ICU patients, collected over 13 years across 10 wards.
- 2Reports were created by 134 radiologists using a standardized template covering eight pathologies and 5-point severity grading.
- 3Model built using this data achieved AUROC scores of 0.80-0.91 for various findings, e.g., 0.91 for right pleural effusion.
- 4Severity grading accuracy (Cohen's kappa) ranged from 0.55 to 0.69.
- 5Most common findings: mild atelectasis (>60% left-sided), mild pulmonary congestion (45%), and enlarged heart (47% of cardiac assessments).
- 6Dataset and code are publicly available via HuggingFace and GitHub.
Why It Matters
Large, high-quality, and openly-shared datasets are critical for advancing AI model development in diagnostic radiology. TAIX-Ray can accelerate research, benchmark progress, and improve the clinical relevance of chest x-ray AI tools in critical care environments.
Source
AuntMinnie
Related News
•AuntMinnie
Study: Computer Vision Models Best LLMs in Chest CT Breast Abnormality Detection
Computer vision models (CVMs) surpass large language models (LLMs) in accurately labeling incidental breast abnormalities on chest CT scans.
•Cardiovascular Business
Radiology Maintains Lead in FDA-Cleared AI Algorithms, Cardiology Follows
Radiology remains the top specialty for FDA-cleared AI, with cardiology as a strong second, particularly in cardiovascular imaging.
•AuntMinnie
Deep Learning Models Rival Radiologists for Pancreatic Cancer Detection on CT
Deep-learning models achieved comparable or superior accuracy to experienced radiologists in detecting pancreatic cancer on CT scans, especially for small tumors.