Magnetic resonance imaging brain glioma IDH gene prediction method and system

Abstract

The invention relates to the technical field of image segmentation and detection, in particular to a magnetic resonance imaging brain glioma IDH gene prediction method and system, and the method comprises the steps: constructing a brain glioma region prediction network; the prediction network comprises a generation network, a discrimination network and a segmentation network; the generative network is used for self-adaptive conversion between the source domain data set and the target domain data set; training a prediction network by using the source domain and target domain data sets based on the loss objective function; the loss objective function is determined according to adversarial loss, cyclic consistency loss, segmentation loss and segmentation consistency loss; carrying out brain glioma region segmentation on a to-be-predicted magnetic resonance image by adopting the trained brain glioma region prediction network; radiomics and a deep neural network are respectively adopted to extract brain glioma region omics features and depth features after segmentation; the omics features and the depth features are combined to serve as input of a random forest algorithm, and classification of IDH genotypes is output. According to the invention, the robustness of genotype prediction is improved.

Description

technical field [0001] The present invention relates to the technical field of image segmentation and detection, in particular to a magnetic resonance imaging brain glioma IDH gene prediction method and system. Background technique [0002] Glioma is one of the most common and malignant brain diseases, and its occurrence and development are closely related to the status of certain genes. One of the most obvious gene mutations is related to isocitrate dehydrogenase (IDH), which can be used as a biomarker for glioma diagnosis and grading, and has important clinical significance for patient treatment planning and prognosis assessment. However, IDH mutation status is usually obtained by histopathological biopsy, which is an invasive and biased diagnostic modality. [0003] Multiparametric MRI (Magnetic Resonance Imaging, Magnetic Resonance Imaging) data provide rich contrast and mutual information from the perspective of imaging phenotypes. However, similar morphologies betwee...

AI Technical Summary

Problems solved by technology

However, there are some obvious defects in this method: first, it is necessary to find a completely consistent data set published on the Internet for migration, which is very demanding for new tasks; The difference between the network public data sets and clinical data for the same task still has data set deviations, which makes the transfer learning results poor.

Images