Attention-guided deep learning for gestational age prediction using fetal brain MRI

Liyue Shen,Edward H Lee,Beth M Kline-Fath,Kristen W Yeom,Lillian M Lai,Jane Chueh,Hisham Dahmoush,John M Pauly,Katie Shpanskaya,Jimmy Zheng,Emily S Mckenna,Dinko Plasto,Mahesh G Atluri,Safwan S Halabi,Courtney Mitchell,Quin Lu,Ozgur Oztekin,Carolina V Guimaraes,Lei Xing

doi:10.1038/s41598-022-05468-5

Abstract

Magnetic resonance imaging offers unrivaled visualization of the fetal brain, forming the basis for establishing age-specific morphologic milestones. However, gauging age-appropriate neural development remains a difficult task due to the constantly changing appearance of the fetal brain, variable image quality, and frequent motion artifacts. Here we present an end-to-end, attention-guided deep learning model that predicts gestational age with R2 score of 0.945, mean absolute error of 6.7 days, and concordance correlation coefficient of 0.970. The convolutional neural network was trained on a heterogeneous dataset of 741 developmentally normal fetal brain images ranging from 19 to 39 weeks in gestational age. We also demonstrate model performance and generalizability using independent datasets from four academic institutions across the U.S. and Turkey with R2 scores of 0.81–0.90 after minimal fine-tuning. The proposed regression algorithm provides an automated machine-enabled tool with the potential to better characterize in utero neurodevelopment and guide real-time gestational age estimation after the first trimester.

Highlights

Magnetic resonance imaging offers unrivaled visualization of the fetal brain, forming the basis for establishing age-specific morphologic milestones
Between the two age prediction approaches, averaging the outputs from the global branch and attention-guided local branch generated higher R 2 scores and smaller mean absolute error (MAE) compared with predictions based on global images alone
Deep learning has emerged as a powerful approach for interpreting complex image features

Summary

Introduction

Magnetic resonance imaging offers unrivaled visualization of the fetal brain, forming the basis for establishing age-specific morphologic milestones. Gauging age-appropriate neural development remains a difficult task due to the constantly changing appearance of the fetal brain, variable image quality, and frequent motion artifacts. The convolutional neural network was trained on a heterogeneous dataset of 741 developmentally normal fetal brain images ranging from 19 to 39 weeks in gestational age. A growing body of neuroscience research has managed to leverage multiple imaging modalities to accurately predict the “brain age” of individuals using machine learning[7,8,9]. These algorithms learn the relationship between neuroimaging features and corresponding ages, after which they are tested on unseen data. Fetal brain MRI protocols, imaging platforms, and operator experience differ widely across institutions, leading to inconsistency in image quality and interpretation[23]

Methods

Results

Discussion

Conclusion