A variant form of 3D-UNet for infant brain segmentation

Saqib Qamar, Hai Jin, Ran Zheng, Parvez Ahmad, Mohd Usama

Abstract: Highlights•We propose a novel 3D CNN architecture with the use of dense connections and inception module for infant brain tissues segmentation. The application of densely connected stage, residual-inception stage, and up-sampling stage achieves excellent results while sustaining computational power.•Simple and adaptable proposed application for Infant brain MRI segmentation.•Most up-to-date results for infant brain segmentation. This suggested approach achieves dice scores of 0.95, 0.905, and 0.92 in CSF, WM, and GM tissues respectively.