CoReFuse-Med: multimodal MRI segmentation network that filters low-quality modalities
Overview
Researchers from Huazhong Normal University and collaborators present CoReFuse-Med, an MRI segmentation network that filters low-quality modalities before fusing T1, T2 and FLAIR images.
The paper reports Dice scores of 0.7326 on EPVS, 0.8528 on BraTS and 0.80 on WMH, with 2.665M parameters on WMH, according to Leiphone (雷峰网). The authors say the method targets resampling artifacts in low-resolution sequences. They say it does not cover misaligned or fully missing modalities.
Written by AI from the articles below · updated Oct 19, 9:55 AM ET
Check the sources:
Article timeline
The articles in this story. Times are ET.
- LeiphoneNewsCoReFuse-Med cleans low-quality MRI modalities before fusion for brain segmentation
AIHuazhong Normal University and collaborators present CoReFuse-Med, a multimodal medical image segmentation network for MRI that filters low-quality modalities before fusing T1, T2 and FLAIR. The paper reports DSC of 0.7326 on EPVS, 0.8528 on BraTS and 0.80 on WMH, with 2.665M parameters on WMH. The authors say the work addresses resampling artifacts in low-resolution sequences and does not cover misaligned or fully missing modalities.
Heat Trend
Not enough continuous observations to show a trend yet.