Blood Vessel Wall Extraction Using Deep Learning Segmentation

The patent applies segmentation by dividing the blood vessel wall into distinct regions using deep learning networks. The system segments the blood vessel wall from surrounding tissues and further divides it into regions containing plaques and regions without plaques, enabling accurate identification of plaque locations despite tissue similarity.

The patent introduces an intermediary approach by using deep learning networks as a mediator between the raw CTA images and plaque analysis. The neural network learns to distinguish plaque regions from non-plaque regions within the blood vessel wall, acting as an intelligent intermediary that overcomes the limitation of traditional image processing methods.

The patent applies universality by designing a multi-functional deep learning network that performs multiple tasks: segmenting the blood vessel wall, identifying plaque regions, and extracting relevant features. This single unified model handles multiple processing requirements, reducing overall system complexity despite the sophisticated processing needed.

The patent applies preliminary action by pre-training the deep learning network on large datasets of CTA images before deployment. The network is preliminarily trained to recognize various tissue patterns and plaque characteristics, so that when actual plaque analysis is needed, the pre-trained model can quickly and accurately process images without requiring complex real-time processing.