Automated Cerebral Infarct Detection on Computed Tomography Images Based on Deep Learning

Area of Science:

• Medical Imaging
• Artificial Intelligence
• Neurology

Background:

• Cerebral infarct detection on CT images is limited by low contrast, hindering its use for early ischemic stroke screening.
• Accurate and timely diagnosis is crucial for effective stroke management and patient outcomes.

Purpose of the Study:

• To enhance the accuracy of automated cerebral infarct detection on CT images using a convolutional neural network (CNN).
• To establish CT as a more reliable first-line diagnostic modality for cerebral infarct screening.

Main Methods:

• CT images were preprocessed to enhance parenchymal contrast, normalize orientation, and create patient-specific t-score maps.
• A CNN was trained using 16x16 pixel patches from t-score maps, with data augmentation applied to both infarcted and non-infarcted patches.
• The CNN model processed t-score matrices as input for patch-wise classification.

Main Results:

• The proposed CNN achieved a patch-wise detection accuracy of 93.9% on the test dataset.
• The method demonstrated prompt and accurate identification of cerebral infarcts on CT images.

Conclusions:

• The developed CNN significantly improves cerebral infarct detection accuracy on CT scans.
• This AI-driven approach supports CT as a viable frontline tool for emergent and regular ischemic stroke detection.