News

Researchers develop a cutting-edge deep-learning model that achieves high accuracy in liver tumor segmentation with a limited training dataset

Signaling a major advance in small-data AI for medical imaging, a research team from Institute of Science Tokyo has developed a novel artificial intelligence model called “MHP-Net,” which delivers cutting-edge performance in liver tumor segmentation with a limited dataset. Unlike conventional models that require thousands of patient cases, this model uses sampling of 3D patches obtained from 7 to 28 tumors. Also, it outperforms the top entry of the MICCAI 2017 (worldwide liver tumor segmentation competition).
MHP-Net: A Smarter AI Tool for Liver Tumor Assessment

Contact

Prof. Kenji Suzuki（Applied Artificial Intelligence Research Core）
https://www.bmai.iir.titech.ac.jp/