Segmentation of Astrocytoma using mutual-attention multimodal MRI

• Published in: Transactions of Japanese Society for Medical and Biological Engineering Vol. Annual62; no. Proc; pp. 403 - 405
• Main Authors: 瀬下, 博之, ラシド, イサム
• Format: Journal Article
• Language: Japanese
• Published: 公益社団法人 日本生体医工学会 2024; Japanese Society for Medical and Biological Engineering
• ISSN: 1347-443X; 1881-4379
• DOI: 10.11239/jsmbe.Annual62.403

Astrocytoma is one of common brain tumors that is known to be aggressive in high grades. Early detection and characterization would enable improvement of patient survival rate. MRI is the leading imaging technology in brain tumor diagnosis, as it provides high contrast imaging with a variety of imaging sequencing. In this study, we analyze the use of multi-modality MRI sequences in automatic segmentation of astrocytoma. Moreover, a mutual-attention deep learning model is proposed for enhancing segmentation accuracy using multimodal MRI. The proposed method is tested using UCSF-PDGM dataset (with 35 astrocytoma subjects). Results indicate that T2 and T2/FLAIR are the most significant MRI modalities. Mutual attention segmentation can provide segmentation accuracy with average dice coefficient of 82%.