B-type aortic dissection CT image segmentation method based on semi-supervised deep learning
CN122990192APending Publication Date: 2026-10-27NINGXIA HUI AUTONOMOUS REGION PEOPLES HOSPITAL
View PDF 0 Cites -1 Cited by
Patent Information
- Application Number
- CN202611028527.4
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2026-07-10
- Publication Date
- 2026-10-27
AI Technical Summary
Technical Problem
[0003]本申请实施例的目的在于提供基于半监督深度学习的B型主动脉夹层CT影像分割方法，用以解决传统卷积网络对B型主动脉夹层的分割结果往往较差的问题
Benefits of technology
[0016]本申请实施例有益效果：本申请实施例提供了基于半监督深度学习的B型主动脉夹层CT影像分割方法，所述方法包括：获取样本CTA影像；将所述样本CTA影像输入待训练的教师学生网络，得到教师网络和学生网络分别对应的影像分割结果，其中，所述待训练的教师学生网络包括教师网络和学生网络；对比所述教师网络的标注结果和所述学生网络对应的影像分割结果，并根据对比结果计算所述学生网络的当前损失；根据所述当前损失对所述学生网络的参数进行调整，并返回所述对比所述教师网络的标注结果和所述学生网络对应的影像分割结果，并根据对比结果计算所述学生网络的当前损失的步骤继续进行训练，直至满足预设迭代停止条件，得到训练好的检测模型；获取待检测CTA影像，并将所述待检测CTA影像输入所述训练好的检测模型，得到所述待检测CTA影像的分割结果，其中，所述待检测CTA影像的分割结果包括：真腔、假腔、假腔血栓中的一种或多种。通过本申请的方案，可以利用教师学生网络半监督深度学习的特点，进行真腔、假腔、假腔血栓的分割，提高分割精度，解决传统卷积网络对B型主动脉夹层的分割结果往往较差的问题。
✦ Generated by Eureka AI based on patent content.
Smart Images
Figure CN122990192A_ABST
Abstract
The embodiment of the application provides a B-type aortic dissection CT image segmentation method based on semi-supervised deep learning, comprising: inputting the sample CTA image into a teacher-student network to be trained to obtain image segmentation results corresponding to the teacher network and the student network respectively; comparing the labeled result of the teacher network with the image segmentation result corresponding to the student network, and calculating the current loss of the student network according to the comparison result; adjusting the parameters of the student network according to the current loss to obtain a trained detection model; obtaining a CTA image to be detected, and inputting the CTA image to be detected into the trained detection model to obtain a segmentation result of the CTA image to be detected, wherein the segmentation result of the CTA image to be detected comprises one or more of a true lumen, a false lumen and a false lumen thrombus. The problem that the segmentation result of the traditional convolutional network for the B-type aortic dissection is often poor can be solved.
Need to check novelty before this filing date? Find Prior Art