Stroke Lesion Segmentation with Visual Cortex Anatomy Alike Neural Nets
This work addresses the need for fast and precise stroke lesion detection, which is crucial for diagnosis and treatment, but it appears incremental as it builds on existing CNN-based methods with a novel anatomical alignment.
The authors tackled stroke lesion segmentation by proposing a neural network model that mimics the human visual cortex anatomy, achieving performance equal to or better than the standard U-Net in preliminary experiments.
Cerebrovascular accident, or commonly known as stroke, is an acute disease with extreme impact on patients and healthcare systems and is the second largest cause of death worldwide. Fast and precise stroke lesion detection and location is an extreme important process with regards to stroke diagnosis, treatment, and prognosis. Except from the manual segmentation approach, machine learning based segmentation methods are the most promising ones when considering efficiency and accuracy, and convolutional neural network based models are the first of its kind. However, most of these neural network models do not really align with the brain anatomical structures. Intuitively, this work presents a more brain alike model which mimics the anatomical structure of the human visual cortex. Through the preliminary experiments on the stroke lesion segmentation task, the proposed model is found to be able to perform equally well or better to the de-facto standard U-Net. Part of the implementation will be made available at https://github.com/DarkoBomer/VCA-Net.