This project is an extension of the efforts to leverage deep learning to detect large vessel occlusion (LVO) from CT angiography. It is motivated by the fact that tedious manual windowing of the CTA by domain experts is often required for the LVO identification algorithm to be successful.
With the recorded window
width and window level adjustment, we are able to train a multi-task neural network to predict the ideal windowing. A demo picture is attached below to illustrate
the effectiveness of our method.
You can contact Sneha (sneha.lingam@vanderbilt.edu) with any questions or ideas regarding the LVO detection project.