Related Experiment Video
Updated: Jun 25, 2025
09:59
A Magnetic Resonance Imaging Protocol for Stroke Onset Time Estimation in Permanent Cerebral Ischemia
Published on: September 16, 2017
14.1K
Deep learning-based magnetic resonance imaging analysis for chronic cerebral hypoperfusion risk
Meiyi Yang1,2, Lili Yang3, Qi Zhang3
1Quzhou Affiliated Hospital of Wenzhou Medical University, Quzhou People's Hospital, Quzhou, Zhejiang, China.
Medical Physics
|May 31, 2024
Summary
A novel deep learning model, CCH-Network (CCHNet), accurately diagnoses chronic cerebral hypoperfusion (CCH) using MRI. This AI tool shows high sensitivity and specificity, improving diagnostic accuracy for CCH and aiding clinical decisions.
Area of Science:
- Medical Imaging
- Artificial Intelligence in Medicine
- Neurology
Background:
- Chronic cerebral hypoperfusion (CCH) presents diagnostic challenges due to nonspecific symptoms.
- Accurate diagnosis of CCH is crucial for effective patient management.
Purpose of the Study:
- To enhance the diagnosis of CCH using Magnetic Resonance Imaging (MRI).
- To improve clinical decision-making and patient treatment outcomes through elevated diagnostic accuracy.
Main Methods:
- A retrospective study utilized 204 routine brain MRIs for training and testing, with 108 for validation.
- Developed CCH-Network (CCHNet), a deep learning model integrating convolution and Transformer modules for structural information capture.
- Employed an adversarial training method to enhance feature knowledge, generalization, and efficiency in CCH risk prediction.
Main Results:
- CCHNet achieved 91.6% AUC and 85.0% accuracy in the testing cohort.
- The model demonstrated 80.0% sensitivity and 90.0% specificity in the testing cohort.
- Validation cohort results showed 86.0% AUC, 84.2% accuracy, 83.3% sensitivity, and 84.7% specificity.
Conclusions:
- CCHNet significantly improved MRI diagnostic performance for CCH.
- The model offers high sensitivity and specificity, presenting a promising new method for CCH diagnosis.
