Probabilistic Learning Coherent Point Drift for 3D Ultrasound Fetal Head Registration

Abstract

Quantification of brain growth is crucial for the assessment of fetal well being, for which ultrasound (US) images are the chosen clinical modality. However, they present artefacts, such as acoustic occlusion, especially after the 18th gestational week, when cranial calcification appears. Fetal US volume registration is useful in one or all of the following cases: to monitor the evolution of fetometry indicators, to segment different structures using a fetal brain atlas, and to align and combine multiple fetal brain acquisitions. This paper presents a new approach for automatic registration of real 3D US fetal brain volumes, volumes that contain a considerable degree of occlusion artefacts, noise, and missing data. To achieve this, a novel variant of the coherent point drift method is proposed. This work employs supervised learning to segment and conform a point cloud automatically and to estimate their subsequent weight factors. These factors are obtained by a random forest-based classification and are used to appropriately assign nonuniform membership probability values of a Gaussian mixture model. These characteristics allow for the automatic registration of 3D US fetal brain volumes with occlusions and multiplicative noise, without needing an initial point cloud. Compared to other intensity and geometry-based algorithms, the proposed method achieves an error reduction of 7.4% to 60.7%, with a target registration error of only 6.38 ± 3.24 mm. This makes the herein proposed approach highly suitable for 3D automatic registration of fetal head US volumes, an approach which can be useful to monitor fetal growth, segment several brain structures, or even compound multiple acquisitions taken from different projections.

Automatic measurements of fetal intracranial volume from 3D ultrasound scans

“…For example, two-stage convolutional neural network (CNN) was used to obtain a full segmentation of the skull (Cerrolaza et al, 2018 ). In another example, a geometric-based feature detection using the point-drift method and random forest tree was used to register different scans of the same subject to each other (Perez-Gonzalez et al, 2020 ). Similarly, after manual brain editing, a constitutional regression network was used to estimate brain age based on its folding program (Namburete et al, 2017 ).…”

Recent Advances in Machine Learning Applied to Ultrasound Imaging

“…Ultrasound imaging was introduced into the field of obstetrics by Donald et al [158], and, since then, it has become the most commonly used imaging modality for investigating several factors related to fetal diagnosis, such as information on fetal biometric measurements, including head and abdominal circumferences, biparietal diameter and information on fetal cardiac activity. Several scientific studies have been devoted to advancement of the quality of prenatal diagnoses by focusing on three main issues: detection of anomalies, fetal measurements, scanning planes and heartbeat [99,100,[159][160][161][162][163][164], segmentation of fetal anatomy in ultrasound images and videos [99,[130][131][132][164][165][166][167] and classification of fetal standard planes, congenital anomalies, biometric measures, and fetal facial expressions [99,100,163,165,[167][168][169][170][171][172][173].…”

Automated segmentation of fetal intracranial volume in 3D ultrasound using deep learning: identifying sex differences in prenatal brain development

“…Recent work has demonstrated the remarkable performance of deep learning techniques on automated biometrics measurement and segmentation of brain structures in both 2D and 3D ultrasound images (13)(14)(15)(16)(17)(18)(19)(20)(21)(22)(23)(24). Moreover, measurements generated in near real time by deep learning techniques could speed up clinical workflow.…”