Fetal Pose Estimation in Volumetric MRI Using a 3D Convolution Neural Network

Abstract

The performance and diagnostic utility of magnetic resonance imaging (MRI) in pregnancy is fundamentally constrained by fetal motion. Motion of the fetus, which is unpredictable and rapid on the scale of conventional imaging times, limits the set of viable acquisition techniques to single-shot imaging with severe compromises in signal-tonoise ratio and diagnostic contrast, and frequently results in unacceptable image quality. Surprisingly little is known about the characteristics of fetal motion during MRI and here we propose and demonstrate methods that exploit a growing repository of MRI observations of the gravid abdomen that are acquired at low spatial resolution but relatively high temporal resolution and over long durations (10-30 minutes). We estimate fetal pose per frame in MRI volumes of the pregnant abdomen via deep learning algorithms that detect key fetal landmarks. Evaluation of the proposed method shows that our framework achieves quantitatively an average error of 4.47 mm and 96.4% accuracy (with error less than 10 mm). Fetal pose estimation in MRI time series yields novel means of quantifying fetal movements in health and disease, and enables the learning of kinematic models that may enhance prospective mitigation of fetal motion artifacts during MRI acquisition.

Rapid head‐pose detection for automated slice prescription of fetal‐brain MRI

“…[26][27][28][29] Segmentation CNNs build on a U-Net architecture 30 and train on relatively large datasets, although augmentation strategies have been used to expose models to more variability than the training data encompasses. 31,32 For many fetal applications, training requires manual segmentations 27 or other ground-truth information 33 to be available. Recent advances employ convolutional encoders to estimate the fetal-head pose from HASTE-slice stacks and orient individual image slices within a template volume.…”

“…26 , 27 , 28 , 29 Segmentation CNNs build on a U‐Net architecture 30 and train on relatively large datasets, although augmentation strategies have been used to expose models to more variability than the training data encompasses. 31 , 32 For many fetal applications, training requires manual segmentations 27 or other ground‐truth information 33 to be available.…”

ENRICH: Exploiting Image Similarity to Maximize Efficient Machine Learning in Medical Imaging

“…Deep learning (DL) has been applied with success in proofs of concept across biomedical imaging modalities and medical specialties [1][2][3][4][5][6][7][8][9][10][11][12][13][14][15][16][17] . DL models can classify images by disease or structure and can segment, track, and measure structures within images.…”

Enhanced Detection of Fetal Pose in 3D MRI by Deep Reinforcement Learning with Physical Structure Priors on Anatomy

“…It's time consuming to manually annotate landmarks of fetal pose, and early demonstrations by Xu et al [19] have addressed this problem with deep learning in convolution neural networks (CNN). Deep reinforcement learning (DRL) [13] is a candidate for an alternative and powerful tool to handle this task.…”