Artificial intelligence applications for the diagnosis and risk evaluation of cerebrovascular disease have been an active research area in recent years. Deep learning-based models have shown promising performance in detecting intracranial aneurysms and large vessel occlusions from a wide range of brain imaging. Feature-based machine learning methods have also successfully evaluated aneurysm rupture risk and stroke reoccurrence. However, to integrate artificial intelligence into real clinical scenarios, several challenges still need to be solved. First, current methods are designed to detect a certain kind of disease, which limits the use of such models in the real-world screening and diagnosis working pipeline. A versatile model that can see various cerebrovascular lesions is needed to suit the current clinical practice. Second, deep learning models still lack generalizability and robustness when models trained on a limited dataset are applied to analyze data collected from different scanners on different sites. Scanner-neutral and site-neutral deep learning methods are desired if a model is to be promoted to multiple centers. Third, current deep learning models lack interpretability, reducing the neurosurgeons' confidence in using them. Explainable artificial intelligence could be an emerging solution to unveil the black-box nature of deep learning-based models.
This Research Topic aims to present a collection of articles that help address the above challenges faced in applying artificial intelligence to cerebrovascular diseases. Cerebrovascular disease is the most common life-threatening neurological event and is the second leading cause of death worldwide. Artificial intelligence has the potential to aid the early diagnosis of cerebrovascular disease and provide better risk management for patients. This Research Topic will cover recent developments in machine learning and deep learning methods for detecting cerebrovascular disease using imaging and clinical data. Methods that attempt to improve the versatility of algorithms in the early screening of cerebrovascular diseases are highly supported. This Research Topic also aims to investigate how to improve artificial intelligence's generalizability and robustness in diagnosing cerebrovascular diseases. We are also interested in methods that will enhance deep learning models' interpretability and better integration of human and machine intelligence.
Potential topics include but are not limited to the following:
- Simultaneous detection and screening of multiple cerebrovascular diseases
- Multimodal data integration for cerebrovascular disease diagnosis
- Explainable artificial intelligence for risk assessment of cerebrovascular diseases
- Robust and generalizable artificial intelligence models for cerebrovascular diseases
- Novel computational methods and integration with artificial intelligence for cerebrovascular hemodynamic analysis
- Novel imaging biomarker discovery using artificial intelligence
- Novel imaging processing to aid cerebrovascular disease analysis
- Neurointerventional education and skill improvement
Artificial intelligence applications for the diagnosis and risk evaluation of cerebrovascular disease have been an active research area in recent years. Deep learning-based models have shown promising performance in detecting intracranial aneurysms and large vessel occlusions from a wide range of brain imaging. Feature-based machine learning methods have also successfully evaluated aneurysm rupture risk and stroke reoccurrence. However, to integrate artificial intelligence into real clinical scenarios, several challenges still need to be solved. First, current methods are designed to detect a certain kind of disease, which limits the use of such models in the real-world screening and diagnosis working pipeline. A versatile model that can see various cerebrovascular lesions is needed to suit the current clinical practice. Second, deep learning models still lack generalizability and robustness when models trained on a limited dataset are applied to analyze data collected from different scanners on different sites. Scanner-neutral and site-neutral deep learning methods are desired if a model is to be promoted to multiple centers. Third, current deep learning models lack interpretability, reducing the neurosurgeons' confidence in using them. Explainable artificial intelligence could be an emerging solution to unveil the black-box nature of deep learning-based models.
This Research Topic aims to present a collection of articles that help address the above challenges faced in applying artificial intelligence to cerebrovascular diseases. Cerebrovascular disease is the most common life-threatening neurological event and is the second leading cause of death worldwide. Artificial intelligence has the potential to aid the early diagnosis of cerebrovascular disease and provide better risk management for patients. This Research Topic will cover recent developments in machine learning and deep learning methods for detecting cerebrovascular disease using imaging and clinical data. Methods that attempt to improve the versatility of algorithms in the early screening of cerebrovascular diseases are highly supported. This Research Topic also aims to investigate how to improve artificial intelligence's generalizability and robustness in diagnosing cerebrovascular diseases. We are also interested in methods that will enhance deep learning models' interpretability and better integration of human and machine intelligence.
Potential topics include but are not limited to the following:
- Simultaneous detection and screening of multiple cerebrovascular diseases
- Multimodal data integration for cerebrovascular disease diagnosis
- Explainable artificial intelligence for risk assessment of cerebrovascular diseases
- Robust and generalizable artificial intelligence models for cerebrovascular diseases
- Novel computational methods and integration with artificial intelligence for cerebrovascular hemodynamic analysis
- Novel imaging biomarker discovery using artificial intelligence
- Novel imaging processing to aid cerebrovascular disease analysis
- Neurointerventional education and skill improvement