Intracranial aneurysm detection method based on multi-dimensional feature fusion

Abstract

The invention belongs to the technical field of intracranial aneurysm detection, and particularly relates to an intracranial aneurysm detection method based on multi-dimensional feature fusion. The method comprises the following steps: S1, collecting data, dividing the collected data into three data sets: a training set, a verification set and a test set, wherein the training set and the verification set are used for a training stage of a model, and the test set is used for verifying the performance of the model; S2, screening and preprocessing of data: before model training, performing screening and preprocessing on an original CTA examination report directly obtained from a hospital so as to reduce the influence of irrelevant reports and backgrounds on a segmentation result; S3, constructing and training the model, and designing an H-AttResUNet mixed dimension convolutional neural network model. According to the intracranial aneurysm detection method based on multi-dimensional feature fusion, the intra-slice features and the three-dimensional context features can be effectively detected, and more accurate segmentation of the intracranial aneurysm is realized.

Description

technical field [0001] The invention relates to the technical field of intracranial aneurysm detection, in particular to a multidimensional feature fusion intracranial aneurysm detection method. Background technique [0002] Intracranial aneurysm is a kind of cerebrovascular disease, and the prevalence rate is about 1%-7%. Once the intracranial aneurysm ruptures, the disability and fatality rate will be above 60%. Therefore, early and accurate detection of aneurysms is critical. Digital angiography is the gold standard for the clinical diagnosis of intracranial aneurysms. However, digital angiography is not only an invasive examination with complex operations, but also cannot fully display the relationship between the aneurysm and the surrounding brain tissue. At present, computed tomography angiography is the most commonly used method for early detection of intracranial aneurysms, and it is necessary for clinicians to judge whether a lesion is included based on CTA images...

AI Technical Summary

Problems solved by technology

In recent years, fully convolutional neural networks (including 2D and 3D convolutional neural networks) have achieved remarkable results in medical image segmentation tasks. However, 2D convolution cannot make full use of 3D spatial information, and 3D convolution has high computational costs. GPU high memory consumption

Images