IDH mutation status prediction by a radiomics associated modality attention network

• Published in: The Visual computer Vol. 39; no. 6; pp. 2367 - 2379
• Main Authors: Zhang, Xinran, Shi, Xiaoyu, Iwamoto, Yutaro, Cheng, Jingliang, Bai, Jie, Zhao, Guohua, Han, Xian-hua, Chen, Yen-Wei
• Format: Journal Article
• Language: English
• Published: Berlin/Heidelberg Springer Berlin Heidelberg 01-06-2023; Springer Nature B.V
• Subjects: Artificial Intelligence; Automation; Brain cancer; Classification; Computer Graphics; Computer Science; Datasets; Deep learning; Image Processing and Computer Vision; Magnetic resonance imaging; Medical imaging; Methods; Mutation; Original Article; Radiomics; Weight reduction; Glioma; Isocitrate dehydrogenase; Modality attention; Self-attention; Radiomics
• ISSN: 0178-2789; 1432-2315
• DOI: 10.1007/s00371-022-02452-y

Isocitrate dehydrogenase (IDH) status is an important factor for the diagnosis of gliomas reported in the 2016 World Health Organization classification scheme for gliomas. There is a strong relationship between IDH mutation status and prognosis. The preoperative prediction of IDH status is necessary for appropriate treatment planning. However, existing methods cannot predict IDH status accurately before the operation. In this paper, we propose a radiomics associated modality attention network to predict IDH mutation status on multi-modality MRI images. Our method first predicts the importance of each modality for the classification task and calculates weights, then uses weighted images for prediction. We also present a light-weight and high-performance self-attention network for gliomas tumor classification to solve the overfitting problem. Additionally, we associate radiomics features for computation of modality attention and classification to enhance the classification accuracy. Our method achieves a 0.7246 F1 Score on our private dataset provided by the First Affiliated Hospital of Zhengzhou University (FHZU), which is better than state-of-the-art methods.