Deep neural network-based cerebral hemorrhage diagnosis system
View Patent ↗Provided is a deep neural network-based cerebral hemorrhage diagnosis system including: an input unit which receives a CT image and presents bleeding areas and suspected bleeding areas; a bleeding size classification unit which is provided with the bleeding areas and suspected bleeding areas presented by the input unit and classifies the bleeding areas and suspected bleeding areas by size; a decoding unit which decodes the bleeding areas and suspected bleeding areas, classified by size by the bleeding size classification unit, by applying neural networks of different depth according to the size of the bleeding areas and suspected bleeding areas; and an output unit which sums and outputs, as a final bleeding area, the results decoded according to size by the decoding unit.
1 . A deep neural network-based cerebral hemorrhage diagnosis system, comprising:
an input processor configured to receive a CT image and identify a bleeding area and a suspected bleeding area;
a bleeding area size classification processor configured to receive the bleeding area and the suspected bleeding area identified by the input processor and classify the received bleeding area and the suspected bleeding area by size;
a decoding processor configured to apply neural networks having different depths to the bleeding area and the suspected bleeding area classified by size by the bleeding area size classification processor to perform decoding; and
an output processor configured to combine results decoded by size by the decoding processor and output the combined decoded results as a final bleeding area.
2 . The deep neural network-based cerebral hemorrhage diagnosis system of claim 1 , wherein the CT image is transmitted to the input processor through a bleeding and suspected bleeding area detection network configured based on a U-Net model.
3 . The deep neural network-based cerebral hemorrhage diagnosis system of claim 2 , wherein when applying a kernel to the CT image, the bleeding and suspected bleeding area detection network uses a center surround difference method to find a portion having a difference greater than a predetermined value by comparing a central portion and a peripheral portion of the kernel.
4 . The deep neural network-based cerebral hemorrhage diagnosis system of claim 1 , wherein the input processor focuses on the bleeding area and the suspected bleeding area and identifies the focused bleeding area and the suspected bleeding area using a form of a box.
5 . The deep neural network-based cerebral hemorrhage diagnosis system of claim 4 , wherein the bleeding area size classification processor classifies the bleeding area and the suspected bleeding area by size based on a size of the box focused on the bleeding area and the suspected bleeding area.
6 . The deep neural network-based cerebral hemorrhage diagnosis system of claim 1 , wherein, when outputting the combined decoded results of the final bleeding area, the output processor classifies a class based on the combined decoded results for an entire bleeding area from the decoding processor, generates a bleeding area box, and applies the class and the bleeding area box to the final bleeding area to output the final bleeding area.