Table 1 Summary of dataset, tasks and respective performance metrics.

From: Fetal Assessment Suite (FetAS): a web-based platform for automatic fetal MRI analysis using AI

Task

Dataset

Method

Test Metrics (average)

Artifact Detection

31, 3D SSFP T2 fetal MRIs

Multi-label Classification of artifact and severity per slice (CNN)

Accuracy: 90.3%

Motion Correction

204*, 3D SSFP T2 fetal MRIs

Image-to-Image motion correction (GAN)

SSIM: 93.6%

PSNR: 33.5dB

Anatomy Segmentation

58, 3D SSFP T2 fetal MRIs

Multi-label segmentation per volume (CNN)

Dice†:

Fetal body = 95.06% Amniotic Fluid = 95.50% Placenta = 88.27%

Fetal Orientation

144, 3D SSFP T2 fetal MRIs

Multi-label classification of fetal orientation (CNN)

Accuracy: 82.2%

Placenta Classification

77, 3D SSFP and HASTE T2 fetal MRIs

Classification of placenta previa (CNN)

Accuracy†: 94.84%

  1. * Dataset sizes shown correspond to the versions described in published work; larger internal datasets are used in ongoing testing and model development.
  2. † Metrics shown are preliminary and included for completeness; detailed descriptions of these methods will be reported in forthcoming publications.
Back to article page