SGPNet: A Three-Dimensional Multitask Residual Framework for Segmentation and IDH Genotype Prediction of Gliomas

Glioma is the main type of malignant brain tumor in adults, and status isocitrate dehydrogenase (IDH) mutation highly affects diagnosis, treatment, prognosis gliomas. Radiographic medical imaging provides a noninvasive platform for sampling both inter intralesion heterogeneity gliomas, previous research has shown that IDH genotype can be predicted from fusion multimodality radiology images. The features images are vital treatment; however, it still lacks multitask framework segmentation lesion areas gliomas prediction genotype. In this paper, we propose novel three-dimensional (3D) deep learning model (SGPNet). residual units also introduced into SGPNet allows output blocks to extract hierarchical different tasks facilitate information propagation. Our reduces 26.6% classification error rates comparing with models on datasets Multimodal Brain Tumor Segmentation Challenge (BRATS) 2020 Cancer Genome Atlas (TCGA) gliomas’ databases. Furthermore, first practically investigate influence performance by setting groups targets. experimental results indicate more important prediction. effective generalizable, which serve as automated tool applied clinical decision making.