Multi-view Attention for gestational age at birth prediction

TL;DR

This paper introduces a multi-view shape analysis method using 2D CNNs and attention layers to predict gestational age at birth from brain surface features, achieving low mean absolute error.

Contribution

The novel approach combines multi-view 2D renderings of 3D brain surfaces with CNNs and attention mechanisms for improved gestational age prediction.

Findings

Achieved MAE of 1.637 on native space

Achieved MAE of 1.38 on template space

Demonstrated effectiveness of multi-view attention in neuroimaging

Abstract

We present our method for gestational age at birth prediction for the SLCN (surface learning for clinical neuroimaging) challenge. Our method is based on a multi-view shape analysis technique that captures 2D renderings of a 3D object from different viewpoints. We render the brain features on the surface of the sphere and then the 2D images are analyzed via 2D CNNs and an attention layer for the regression task. The regression task achieves a MAE of 1.637 +- 1.3 on the Native space and MAE of 1.38 +- 1.14 on the template space. The source code for this project is available in our github repository https://github.com/MathieuLeclercq/SLCN_challenge_UNC