Lightweight baby brain tissue image segmentation method based on context information

Abstract

The invention discloses a lightweight infant brain tissue image segmentation method based on context information, which is applied to the technical field of image processing, and comprises the following steps: in a training stage, constructing a convolutional neural network and training to obtain a corresponding segmentation prediction map; a loss function value between a set formed by segmentation prediction maps corresponding to an original nuclear magnetic resonance imaging T1 mode and a set formed by corresponding infant brain tissue distortion segmentation labeling maps is calculated, and an Adam optimizer is used to update network parameters; inputting the nuclear magnetic resonance imaging T1 mode and the corresponding nuclear magnetic resonance imaging T2 mode into a convolutional neural network training model to obtain an infant brain tissue segmentation image; according to the method, the segmentation effect in various brain tissue types is well expressed, the parameter quantity is lower, the weight is smaller, the reasoning speed is higher, and deployment on mobile terminal equipment is facilitated.

Description

technical field [0001] The invention relates to the technical field of image processing, and more particularly to a light-weight infant brain tissue image segmentation method based on context information. Background technique [0002] In recent years, the technologies of artificial intelligence and computer vision are rising rapidly, and with the development of the Internet, a huge amount of data has been brought for practical applications, and the rapid acquisition of key information from the massive image and video data has become the field of computer vision. The key issue. Among them, medical image analysis has attracted more and more attention. Using MRI, images of the baby's brain tissue can be collected. Using these data combined with computer vision technology can make a big difference in the field of life and health. [0003] The ideal intelligent assisted diagnosis and treatment requires a correct visual understanding of brain tissue. Using deep learning method...

AI Technical Summary

Problems solved by technology

[0004] With the emergence of various segmentation models, the final segmentation performance indicators are also rising. However, considering that these segmentation models have relatively large storage capacity and large floating-point calculations, it will be more difficult to deploy them on mobile terminals. big challenge

Images