White matter hyperintensities automatic segmentation system and method based on Unet model

Abstract

The invention discloses a white matter hyperintensitiesl automatic segmentation system and method based on a Unet model, and the system comprises an MRI image preprocessing module which is used for carrying out the normalization of inputted MRI image data, and removing the noise interference; a skull removing module which is used for removing the skull in the image by utilizing a skull removing algorithm so as to remove a non-brain tissue part and further remove noise interference; and a WMH segmentation module which is used for reading the MRI image data processed by the skull removal module,converting the MRI image data into image information data, transmitting the image information data to the neural network model for segmentation result prediction, and outputting an accurate segmentation result. By adopting the method, the WMH region can be automatically segmented by applying a deep learning algorithm in the process of automatically processing the image, so that the white matter hyperintensities focus of magnetic resonance imaging can be quantitatively and efficiently distinguished, and the aims of reducing the workload of doctors and facilitating subsequent diagnosis and research are fulfilled.

Description

technical field [0001] The invention relates to medical imaging and magnetic resonance imaging (MRI) image processing technology, in particular to an automatic white matter hyperintensity (White Matter Hyperintensities, WMH) segmentation system and method based on Unet model. Background technique [0002] Unet, or Unity Networking, is a network communication solution. The Unet model is an image segmentation network built using Unet technology. [0003] The brain is composed of tens of billions of neurons, and neurons are composed of cell bodies and nerve fibers. The cell body has a nucleus (dark color), and the nerve fibers have cytoplasm (light color). In the brain, cell bodies gather on the surface of the brain and are called gray matter because they look dark; while nerve fibers gather inside the brain and look light in color, so they are called white matter. [0004] White matter hyperintensities (WMH) are multiple white matter around the lateral ventricles or subcorte...

AI Technical Summary

Problems solved by technology

[0005] Although there are many automatic segmentation methods for WMH in the current magnetic resonance imaging (MRI) processing technology, in order to obtain clinically practical results, doctors usually need to perform manual correction, especially MRI images often contain dozens to hundreds of layers. Each layer needs to be manually corrected by the doctor. Due to the huge workload, it is easy to cause the doctor's visual fatigue and increase the chance of error

At the same time, due to the various structures of WMH, which appear as irregular polygonal shapes or scattered points, and are randomly distributed, it is difficult for even experienced doctors to quickly make judgments and segment them, so the efficiency is low

Moreover, the challenge of automatic segmentation of WMH not only comes from the variety of MRI equipment, poor imaging quality, uneven signal, random position and irregular distribution of size, MRI noise, artifacts, etc., but also the existence of other brain diseases. The signal enhancement on MRI will further make it difficult to automatically and accurately segment WMH

Images