Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Automatic facial axes standardization of 3D fetal ultrasound images

Antonia Alomar, Ricardo Rubio, Laura Salort, Gerard Albaiges, Antoni Payà, Gemma Piella, Federico Sukno

TL;DR

An AI-driven tool to assist clinicians in standardizing fetal facial axes/planes in 3D US, reducing sonographer workload and facilitating the facial evaluation, demonstrates potential for improving the consistency and accuracy of fetal facial assessments in clinical settings.

Abstract

Craniofacial anomalies indicate early developmental disturbances and are usually linked to many genetic syndromes. Early diagnosis is critical, yet ultrasound (US) examinations often fail to identify these features. This study presents an AI-driven tool to assist clinicians in standardizing fetal facial axes/planes in 3D US, reducing sonographer workload and facilitating the facial evaluation. Our network, structured into three blocks-feature extractor, rotation and translation regression, and spatial transformer-processes three orthogonal 2D slices to estimate the necessary transformations for standardizing the facial planes in the 3D US. These transformations are applied to the original 3D US using a differentiable module (the spatial transformer block), yielding a standardized 3D US and the corresponding 2D facial standard planes. The dataset used consists of 1180 fetal facial 3D US images acquired between weeks 20 and 35 of gestation. Results show that our network considerably reduces inter-observer rotation variability in the test set, with a mean geodesic angle difference of 14.12$^{\circ}$ $\pm$ 18.27$^{\circ}$ and an Euclidean angle error of 7.45$^{\circ}$ $\pm$ 14.88$^{\circ}$. These findings demonstrate the network's ability to effectively standardize facial axes, crucial for consistent fetal facial assessments. In conclusion, the proposed network demonstrates potential for improving the consistency and accuracy of fetal facial assessments in clinical settings, facilitating early evaluation of craniofacial anomalies.

Automatic facial axes standardization of 3D fetal ultrasound images

TL;DR

An AI-driven tool to assist clinicians in standardizing fetal facial axes/planes in 3D US, reducing sonographer workload and facilitating the facial evaluation, demonstrates potential for improving the consistency and accuracy of fetal facial assessments in clinical settings.

Abstract

Craniofacial anomalies indicate early developmental disturbances and are usually linked to many genetic syndromes. Early diagnosis is critical, yet ultrasound (US) examinations often fail to identify these features. This study presents an AI-driven tool to assist clinicians in standardizing fetal facial axes/planes in 3D US, reducing sonographer workload and facilitating the facial evaluation. Our network, structured into three blocks-feature extractor, rotation and translation regression, and spatial transformer-processes three orthogonal 2D slices to estimate the necessary transformations for standardizing the facial planes in the 3D US. These transformations are applied to the original 3D US using a differentiable module (the spatial transformer block), yielding a standardized 3D US and the corresponding 2D facial standard planes. The dataset used consists of 1180 fetal facial 3D US images acquired between weeks 20 and 35 of gestation. Results show that our network considerably reduces inter-observer rotation variability in the test set, with a mean geodesic angle difference of 14.12 18.27 and an Euclidean angle error of 7.45 14.88. These findings demonstrate the network's ability to effectively standardize facial axes, crucial for consistent fetal facial assessments. In conclusion, the proposed network demonstrates potential for improving the consistency and accuracy of fetal facial assessments in clinical settings, facilitating early evaluation of craniofacial anomalies.
Paper Structure (15 sections, 7 equations, 2 figures, 2 tables)

This paper contains 15 sections, 7 equations, 2 figures, 2 tables.

Table of Contents

  1. Introduction
  2. Related Work
  3. Method
  4. Data & Pre-processing:
  5. Ground Truth Standard Planes:
  6. Feature Extractor Block:
  7. Translation & Rotation Regression Block:
  8. Spatial Transformer Block:
  9. Cumulative Transformations & Initialization:
  10. Network Loss:
  11. Experiments
  12. Results & Discussion
  13. Conclusions
  14. Acknowledgments.
  15. Disclosure of Interests.

Figures (2)

  • Figure 1: Proposed architecture. The network is divided in three blocks: feature extractor, rotation and translation regression, and the spatial transformer block. The two first blocks extract the features and estimate the transformation necessary to obtain the standard facial axes/planes taking as input three orthogonal slices in arbitrary position. Then, the differentiable spatial transformer block applies the estimated transformation to the original 3D US to obtain the standardized 3D US and the 2D facial SPs.
  • Figure 2: Qualitative test set examples. Four examples of the estimated 2D planes using the proposed method, the mean Euclidean angle ($\mu_{EA}$) in degrees and the mean translation ($\mu_{t}$) error in mm obtained per plane compared to the GT.