Search

Computerised analysis of 3D ultrasound (US) volumes requires that surface representations are extracted through segmentation. To overcome the limitations of manual segmentation (MS) we developed and atlas-based automatic segmentation (AS) method which exploits a set of manually segmented volumes. The newly developed AS algorithm was trained with 20 and 30 volumes (A 20, A 30) and applied with a leave-one-out approach on 20 volumes. [Extracted from the article]

The inter-operator variability was assessed using Dice scores to evaluate the percentage of overlap between 2 MS. Dice scores between the 3 operators were highly reproducible, with median scores (±IQR) of 92% ±4 (op 1-op 2), 94% ±4 (op 1-op 3) and 94% ±5 (op 2-op 3). [Extracted from the article]

In order to develop computerised analysis of the fetal face, specific facial "landmarks" must be identified and delineated on 3D ultrasound volumes. We sought to assess the feasibility of an automated segmentation technique for quantitative assessment of fetal face three-dimensional ultrasound volumes. [Extracted from the article]