Convolutional Neural Networks to Assess Steno-Occlusive Disease Using Cerebrovascular Reactivity.

Area of Science:

• Neuroimaging
• Artificial Intelligence
• Cardiovascular Medicine

Background:

• Cerebrovascular Reactivity (CVR) is crucial for assessing cerebral blood flow changes using BOLD MRI.
• Steno-occlusive disease (SOD) is a leading cause of ischemic stroke, necessitating improved diagnostic tools.
• Current CVR analysis aids in identifying cerebral perfusion insufficiency.

Purpose of the Study:

• To develop a convolutional neural network (CNN)-based clinical decision support system for SOD patient screening.
• To discriminate between healthy and unhealthy CVR maps for early SOD detection.
• To enhance the early diagnosis and clinical management of cerebrovascular diseases.

Main Methods:

• A CNN model was developed and trained on a dataset of 231 CVR maps (68 healthy, 163 SOD patients).
• Transfer learning experiments were conducted using customized pre-trained networks.
• Image augmentation techniques were applied to the training and validation datasets.

Main Results:

• A customized CNN with a double-stacked convolution layer architecture achieved the best performance.
• The developed system demonstrated results consistent with expert clinical interpretations.
• The CNN effectively discriminated between healthy and SOD CVR maps.

Conclusions:

• CNN-based CVR map analysis offers a promising approach for automated SOD screening.
• This AI tool can support clinicians in the early diagnosis of cerebrovascular insufficiency.
• The study highlights the potential of AI in improving the management of stroke risk factors.