In obstetrics, the application of computer-based image analysis to provide deeper insight into pathology in early pregnancy is highly desirable but underdeveloped. One such pathology, fetal growth restriction (FGR) is a leading cause of mortality and morbidity in pregnancy. FGR affects approximately 3-10% of pregnancies in the western world leading to increased risk of stillbirth and health problems in later life. Morphometric or functional measurement of the placenta in pregnancy in utero may aid diagnosis of this pathology as the interface between placenta and mother is the site where the pathology manifests itself. Detection of growth restriction is yet to be resolved as poor, unreliable biochemical and image-based biomarkers have made it hard to detect and manage these pregnancies effectively. By the provision and development of tools for quantification of the placenta by three dimensional (3D) ultrasound (US) using image segmentation and mesh pro- cessing, this thesis aims to better facilitate clinical investigation of this major problem in obstetric healthcare. In a first contribution, 3D placental volume measurement using 3D US is used to classify the difference between normal and FGR pregnancies. Volume was estim- ated using the semi-automated random walker (RW) algorithm. The repeatability and reliability of the method was tested between three observers and showed the new method to be equivalent to manual segmentation. In a study of 143 women performed by our clinical partners, significantly smaller placental volume was found in pregnancies defined as small-for-gestational age (SGA). Expanding on volumetry, the utero-placental interface (UPI) is the location where the pathology that leads to FGR occurs. Manual manipulation of the volume is requiredtovisualisetheinterface, soweinvestigatedapplyinga“meshflattening” process to convert the contorted UPI into a disc to provide a standardised way to view the interface between acquistions and subjects. Finally, an existing, two dimensional (2D) Doppler standardisation technique was extended into 3D to provide standardisation of values of Doppler vascularity. This technique was then applied to measure the vascularity of volumes of interest relat- ive to the interface between placenta and mother. This test was then applied clin- ically in 143 women and found that the vascularity of the small-for-gestational age (SGA) pregnancies was significantly smaller than that of the population who produced appropriately sized babies. These three tools each provide augmentation of our understanding of placental health and function in pregnancy. From measuring the gross volume to estimat- ing the blood flow we show the potential clinical application for image analysis performed on 3D power Doppler (PD) ultrasound volumes.

<b>Background</b>: Fetal growth restricted (FGR) infants have increased perinatal morbidity and mortality risks. Standard fetal biometry may identify some FGR babies; amniotic fluid measurement and Doppler assessment of blood vessels provide additional, functional assessments, but are often normal in babies with late-onset FGR who are difficult to diagnose. A marker reflecting nutritional status should help to identify FGR fetuses, enabling effective intervention: increased monitoring or delivery and neonatal management. Arm and/or thigh measurements have previously shown potential in 2D and 3D. Limb fat volume has never been measured and could provide an in utero marker of fetal nutritional status. <b>Aims</b>: 1. To develop an ultrasound scanning protocol to acquire 2D images and 3D volumes of fetal arms and thighs. 2. To develop method(s), suitable for use in clinical practice, to measure fat in these images and volumes. 3. To assess the reproducibility of these methods. 4. To assess the validity of these methods by comparing them with MRI images of fetal limb fat. 5. To use these methods in a healthy population to develop reference ranges. 6. To correlate these methods with validated neonatal measurements, to assess whether the antenatal methods reflect neonatal body composition. 7. To apply these methods to longitudinal prospective ultrasound images from multiple countries to assess SGA fetuses. <b>Method</b>: Ultrasound imaging protocols were developed to obtain accurate images and volumes of fetal arms and thighs. Segmentation tools were designed with biomedical engineers to measure fat, lean and limb compartments, and applied to 2D images and 3D volumes (n=500), with calculation of reference centiles in optimally healthy women (INTERGROWTH-21st study) and MRI validation of the ultrasound measurements. Additional methods were assessed: a two-ellipse method, and a three-thickness Fetal Fat Index (FFI). Reproducibility was assessed using Bland-Altman plots and ICCs. Fetal measurements were correlated with neonatal body composition data and anthropometry. Third trimester fetal thigh fat volumes were compared in sub-cohorts of AGA and SGA fetuses. <b>Results</b>: Reference centiles were calculated for novel fractional arm and leg volumes (fat and lean), from 16 to 41 weeks. 2D reference ranges were also calculated. The FFI technique - quick, simple, 2D - correlated well with fat area and fat volume. DXA analyses showed a strong correlation between neonatal limb and whole body fat. Correlation analyses showed that infants with above- and below-average arm circumferences have significantly different amounts of arm fat as early as 30-35 weeks. The strength of correlation between antenatal limb fat and neonatal PEA POD whole body fat increased with increasing gestational age. Scans at 30-34 weeks showed a significant difference in fractional thigh fat between those who would be born SGA compared with AGA. <b>Conclusion</b>: This thesis explores, in detail, the measurement of fetal arm and thigh fat using 2D and 3D ultrasound, and demonstrates that it is correlated to neonatal body composition thus allowing 'fetal body composition' to be established as a research tool; the ultimate aim is to be able to distinguish growth-restricted fetuses from those of normal nutritional status. Novel measurements have been developed, acquisition protocols described, reproducibility assessed, and reference centiles calculated in an optimally healthy population: 2D Fetal Fat Index, 3D fractional limb fat and lean volumes, and limb fat (2D and 3D) as a percentage.