GSGordon N. Stevenson

Gordon N. Stevenson

*This is an algorithmically generated author record based on publicly available information.

Current Organizations / Institutes

The organizations/institutes from the scholar's latest article

Previous Organizations / Institutes

The organizations / institutes from the scholar's other articles in the past ten years


Gordon N. Stevenson has a total of 80 co-authors. Here are the top co-authors by number of shared publications:


Latest Publications

  • Open Access

    Automatic 3D Multi-modal Ultrasound Segmentation of Human Placenta using Fusion Strategies and Deep Learning

    by Sonit Singh, Gordon Stevenson, Brendan Mein, Alec Welsh, Arcot Sowmya

    Purpose: Ultrasound is the most commonly used medical imaging modality for diagnosis and screening in clinical practice. Due to its safety profile, noninvasive nature and portability, ultrasound is the primary imaging modality for fetal assessment in pregnancy. Current ultrasound processing methods are either manual or semi-automatic and are therefore laborious, time-consuming and prone to errors, and automation would go a long way in addressing these challenges. Automated identification of placental changes at earlier gestation could facilitate potential therapies for conditions such as fetal growth restriction and pre-eclampsia that are currently detected only at late gestational age, potentially preventing perinatal morbidity and mortality. Methods: We propose an automatic three-dimensional multi-modal (B-mode and power Doppler) ultrasound segmentation of the human placenta using deep learning combined with different fusion strategies.We collected data containing Bmode and power Doppler ultrasound scans for 400 studies. Results: We evaluated different fusion strategies and state-of-the-art image segmentation networks for placenta segmentation based on standard overlap- and boundary-based metrics. We found that multimodal information in the form of B-mode and power Doppler scans outperform any single modality. Furthermore, we found that B-mode and power Doppler input scans fused at the data level provide the best results with a mean Dice Similarity Coefficient (DSC) of 0.849. Conclusion: We conclude that the multi-modal approach of combining B-mode and power Doppler scans is effective in segmenting the placenta from 3D ultrasound scans in a fully automated manner and is robust to quality variation of the datasets.


  • Fully Automated 3-D Ultrasound Segmentation of the Placenta, Amniotic Fluid, and Fetus for Early Pregnancy Assessment

    by Padraig Looney, Yi Yin, Sally L. Collins, Kypros H. Nicolaides, Walter Plasencia, Malid Molloholli, Stavros Natsis, Gordon N. Stevenson

    Volumetric placental measurement using 3-D ultrasound has proven clinical utility in predicting adverse pregnancy outcomes. However, this metric cannot currently be employed as part of a screening test due to a lack of robust and real-time segmentation tools. We present a multiclass (MC) convolutional neural network (CNN) developed to segment the placenta, amniotic fluid, and fetus. The ground-truth data set consisted of 2093 labeled placental volumes augmented by 300 volumes with placenta, amniotic fluid, and fetus annotated. A two-pathway, hybrid (HB) model using transfer learning, a modified loss function, and exponential average weighting was developed and demonstrated the best performance for placental segmentation (PS), achieving a Dice similarity coefficient (DSC) of 0.84- and 0.38-mm average Hausdorff distances (HDAV). The use of a dual-pathway architecture improved the PS by 0.03 DSC and reduced HDAV by 0.27 mm compared with a naïve MC model. The incorporation of exponential weighting produced a further small improvement in DSC by 0.01 and a reduction of HDAV by 0.44 mm. Per volume inference using the FCNN took 7-8 s. This method should enable clinically relevant morphometric measurements (such as volume and total surface area) to be automatically generated for the placenta, amniotic fluid, and fetus. The ready availability of such metrics makes a population-based screening test for adverse pregnancy outcomes possible.


  • Multi-Parametric Fusion of 3D Power Doppler Ultrasound for Fetal Kidney Segmentation Using Fully Convolutional Neural Networks

    by Nipuna H. Weerasinghe, Nigel Hamilton Lovell, Alec William Welsh, Gordon N. Stevenson

    Kidney development is key to the long-term health of the fetus. Renal volume and vascularity assessed by 3D ultrasound (3D-US) are known markers of wellbeing, however, a lack of real-time image segmentation solutions preclude these measures being used in a busy clinical environment. In this work, we aimed to automate kidney segmentation using fully convolutional neural networks (fCNN). We used multi-parametric input fusion incorporating 3D B-Mode and power Doppler (PD) volumes, aiming to improve segmentation accuracy. Three different fusion strategies and their performance were assessed versus a single input (B-Mode) network. Early input-level fusion provided the best segmentation accuracy with an average Dice similarity coefficient (DSC) of 0.81 and Hausdorff distance (HD) of 8.96 mm, an improvement of 0.06 DSC and reduction of 1.43mm HD compared to our baseline network. Compared to manual segmentation for all models, repeatability was assessed by intra-class correlation coefficients (ICC) indicating good to excellent reproducability (ICC >= 0.93). The framework was extended to support multiple graphics processing units (GPUs) to better handle volumetric data, dense fCNN models, batch normalisation and complex fusion networks. This work and available source code provides a framework to increase the parameter space of encoder-decoder style fCNNs across multiple GPUs and shows that application of multi-parametric 3D-US in fCNN training improves segmentation accuracy.


  • Three-dimensional US Fractional Moving Blood Volume: Validation of Renal Perfusion Quantification

    by Alec W. Welsh, J. Brian Fowlkes, Stephen Z. Pinter, Kimberly A. Ives, Gabe E. Owens, Jonathan M. Rubin, Oliver D. Kripfgans, Pádraig Looney, Sally L. Collins, Gordon N. Stevenson

    With use of three-dimensional power Doppler US, fractional moving blood volume was accurate and reproducible for determining renal perfusion in a porcine model. Background Three-dimensional (3D) fractional moving blood volume (FMBV) derived from 3D power Doppler US has been proposed for noninvasive approximation of perfusion. However, 3D FMBV has never bee...


  • Open Access

    3D ultrasound file reading and coordinate transformations

    by Pádraig Looney, Gordon Stevenson, Sally Collins

    The Kretzfile format is used to store 3D ultrasound data, from GE Voluson ultrasound scanners. The geometry used in these hies is a toroidal coordinate system. Cartesian coordinates are required to allow application of advanced image libraries like ITK and scikit-image.


  • Applying spatial–temporal image correlation to the fetal kidney: Repeatability of 3D segmentation and volumetric impedance indices

    by Bonita Gu, Gordon N. Stevenson, Ana Ferreira, Sudeshni Pathirana, Jennifer Sanderson, Amanda Henry, Jennifer Alphonse, Alec W. Welsh

    Introduction: Spatiotemporal image correlation (STIC) can evaluate fetal renal impedance using four‐dimensional volumetric indices. We assessed repeatability of three‐dimensional kidney segmentation and the repeatability of the resultant indices.Methods: In each of 57 healthy pregnant women, three renal artery pulsed‐wave Doppler (PWD) traces and three STIC volumes were acquired from the same fetal kidney and segmented by two observers. Vascularisation‐flow index (VFI) and fractional moving blood volume (FMBV) were calculated for every STIC frame and used to determine the volumetric pulsatility index (vPI), volumetric resistance index (vRI) and volumetric systolic/diastolic ratio (vS/D). Segmentation performance was assessed using Dice similarity coefficients (DSCs), Hausdorff distances, coefficient of variation (CoV) and the intraclass correlation coefficient (ICC). Intra/Inter volumetric index repeatability was assessed using ICCs.Results: Forty‐eight cases (84%) provided full data. Mean intra‐ and interobserver DSCs were 0.90 and 0.81. Mean intra‐ and interobserver Hausdorff distances were 3.88 mm and 5.27 mm. Average kidney volumes for observers 1 and 2 were 9.88 mL and 8.54 mL (mean difference 16.1%). Mean intra‐observer volumetric CoVs were 5.3% and 8.1%. Intra‐ and interobserver ICCs for kidney volume (same STIC volume) were 0.97 and 0.85. When assessing volume variation between STIC volumes, intra‐observer ICC was 0.97. ICCs were 0.77–0.81 for VFI‐derived volumetric indices and 0.61–0.62 for FMBV‐derived indices; ICCs for all PWD indices were between 0.58 and 0.59.Conclusions: Periodical variation in vascularity was demonstrated in the fetal kidney, and three‐dimensional segmentation was highly repeatable. Derived volumetric impedance indices show moderate variability but outperform corresponding two‐dimensional PWD indices in terms of reproducibility.


  • Open Access

    3D fractional moving blood volume (3D-FMBV) demonstrates decreased first trimester placental vascularity in pre-eclampsia but not the term, small for gestation age baby

    by Sally L. Collins, Alec W. Welsh, Lawrence Impey, J. Alison Noble, Gordon N. Stevenson

    To undertake an observational study to see whether first trimester placental vascularity, measured with a standardized power Doppler index: 3D-FMBV, is different in pregnancies which either develop pre-eclampsia or lead to term, normotensive small for gestational age (SGA) babies. Women were scanned between 11 and 13+6 weeks. The placental volume (sPlaV) was estimated using our previously validated semi-automated tool. Estimates of 3D-FMBV were generated from the raw power Doppler signal for the whole utero-placental interface, UPI (FMBV-UPI) and 5mm into the placenta (FMBV-IVS). Differences in the placental volume and FMBV for pregnancies developing pre-eclampsia and resulting in term, normotensive SGA babies were compared with term, normotensive, appropriate for gestational age (AGA), controls. Results were available for 143 women. The placental volume (sPlaV) was reduced in both pre-eclampsia (p = 0.007) and term, normotensive SGA (p = 0.001) when compared with term normotensive AGA controls. 3D-FMBV estimates were significantly lower for pregnancies developing pre-eclampsia (FMBV-UPI, p = 0.03, FMBV-IVS, p = 0.01) but not for the normotensive SGA pregnancies (FMBV-UPI, p = 0.16, FMBV-IVS, p = 0.27). Pregnancies destined to develop pre-eclampsia are more likely to have small placentas with significantly reduced vascularity at 11–13 weeks. Those pregnancies which were normotensive throughout but resulted in an SGA baby delivered at term, had significantly smaller placentas but with similar vascularity to normotensive AGA pregnancies.


  • Basal ganglia perfusion in the preterm infant during transition

    by Tim Schindler, Yasmin Gilbert, Sonali Jayatilake, Gordon Stevenson, Ju Lee Oei, Alec Welsh

    The preterm brain is susceptible to changes in blood flow. Using power Doppler images, digital imaging techniques have been developed to measure the total amount of blood flow in a defined area, giving the index: fractional moving blood volume (FMBV). The aim of this study was to investigate temporal changes in basal ganglia perfusion during the transitional period after birth. Twenty-four preterm infants were examined with serial cranial ultrasounds at four time points during the first 48 h of life. FMBV was calculated using power Doppler images at each time point. All infants had analyzable data and FMBV was successfully calculated at all time points. Twenty-three of the 24 infants had an increasing trend in FMBV over time. The median FMBV increased from 17% at 6 h to 25% at 48 h. One-way repeated measures ANOVA showed a significant increase in values at P < 0.001 at each of the four time points. We have demonstrated changes in basal ganglia blood flow as the cerebral circulation adapts to extrauterine life. With further investigation, this technique may be useful in the assessment of preterm circulatory adaptation, either alone or in conjunction with other modes of evaluating cerebral blood flow. Download references