Michael Aertsen

GIFT-Cloud

Highlights • A platform for sharing medical imaging data between clinicians and researchers.• Extensible system connects three hospitals and two universities.• Simple for end users with low impact on hospital IT systems.• Automated anonymisation of pixel data and metadata at the clinical site.• Maintains subject data groupings while preserving anonymity.

DeepIGeoS: A Deep Interactive Geodesic Framework for Medical Image Segmentation

Accurate medical image segmentation is essential for diagnosis, surgical planning and many other applications. Convolutional Neural Networks (CNNs) have become the state-of-the-art automatic segmentation methods. However, fully automatic results may still need to be refined to become accurate and robust enough for clinical use. We propose a deep learning-based interactive segmentation method to improve the results obtained by an automatic CNN and to reduce user interactions during refinement for higher accuracy. We use one CNN to obtain an initial automatic segmentation, on which user interactions are added to indicate mis-segmentations. Another CNN takes as input the user interactions with the initial segmentation and gives a refined result. We propose to combine user interactions with CNNs through geodesic distance transforms, and propose a resolution-preserving network that gives a better dense prediction. In addition, we integrate user interactions as hard constraints into a back-propagatable Conditional Random Field. We validated the proposed framework in the context of 2D placenta segmentation from fetal MRI and 3D brain tumor segmentation from FLAIR images. Experimental results show our method achieves a large improvement from automatic CNNs, and obtains comparable and even higher accuracy with fewer user interventions and less time compared with traditional interactive methods.

Tumour-related imaging parameters predicting the percentage of preserved normal renal parenchyma following nephron sparing surgery: a retrospective study

AbstractObjectivesTo examine pre-operative imaging parameters that predict the residual amount of healthy renal parenchyma after nephron sparing surgery (NSS) for renal tumours, as this can help stratify patients towards the optimal surgical choice.MethodsNinety-eight patients with the diagnosis of a solitary unilateral renal tumour and with pre- and post-operative imaging were included in this retrospective study. Imaging, patient and surgical parameters were acquired and their correlation to the percentage decrease of healthy renal parenchyma following surgery was statistically examined to find the most significant predictor of nephron sparing.ResultsLoss of healthy renal parenchyma was highest in patients with renal sinus tumour involvement (P = 0.003) and anterior tumours (P = 0.006), but not significantly correlated with medial/lateral location (P = 0.940) or exophytic/endophytic tumour growth (P = 0.244). The correlation of tumour size with the percentage of parenchymal sparing did not quite reach statistical significance (P = 0.053), but involvement of the urinary collecting system (P = 0.008) was a very good predictor of complications. Loss of healthy renal parenchyma was higher in patients with high-grade surgical complications (P = 0.001).ConclusionsSeveral pre-operative parameters correlate to percentage nephron sparing after NSS. Anterior tumour location and renal sinus involvement proved to be the best predictors of loss of healthy renal parenchyma.Key points• Modern imaging provides exceptional pre-operative information about renal tumours. • Tumour imaging parameters correlate with percentage preserved parenchyma after nephron sparing surgery. • Anterior tumours and sinus involvement strongly predict eventual loss of healthy parenchyma. • This information can contribute to an informed decision about surgical options.

Slic-Seg: Slice-by-Slice Segmentation Propagation of the Placenta in Fetal MRI Using One-Plane Scribbles and Online Learning

Segmentation of the placenta from fetal MRI is critical for planning of fetal surgical procedures. Unfortunately, it is made difficult by poor image quality due to sparse acquisition, inter-slice motion, and the widely varying position and orientation of the placenta between pregnant women. We propose a minimally interactive online learning-based method named Slic-Seg to obtain accurate placenta segmentations from MRI. An online random forest is first trained on data coming from scribbles provided by the user in one single selected start slice. This then forms the basis for a slice-by-slice framework that segments subsequent slices before incorporating them into the training set on the fly. The proposed method was compared with its offline counterpart that is with no retraining, and with two other widely used interactive methods. Experiments show that our method 1) has a high performance in the start slice even in cases where sparse scribbles provided by the user lead to poor results with the competitive approaches, 2) has a robust segmentation in subsequent slices, and 3) results in less variability between users.

Slic-Seg: A minimally interactive segmentation of the placenta from sparse and motion-corrupted fetal MRI in multiple views

Highlights • Minimal user interaction is needed for a good segmentation of the placenta.• Random forests with high level features improved the segmentation.• Higher accuracy than state-of-the-art interactive segmentation methods.• Co-segmentation of multiple volumes outperforms single sparse volume based method.

In vivo evidence by magnetic resonance volumetry of a gestational age dependent response to tracheal occlusion for congenital diaphragmatic hernia

We aimed to assess in vivo changes in lung and liver volumes in fetuses with isolated congenital diaphragmatic hernia, either expectantly managed or treated in utero.

An Automated Localization, Segmentation and Reconstruction Framework for Fetal Brain MRI

Reconstructing a high-resolution (HR) volume from motion-corrupted and sparsely acquired stacks plays an increasing role in fetal brain Magnetic Resonance Imaging (MRI) studies. Existing reconstruction methods are time-consuming and often require user interaction to localize and extract the brain from several stacks of 2D slices. In this paper, we propose a fully automatic framework for fetal brain reconstruction that consists of three stages: (1) brain localization based on a coarse segmentation of a down-sampled input image by a Convolutional Neural Network (CNN), (2) fine segmentation by a second CNN trained with a multi-scale loss function, and (3) novel, single-parameter outlier-robust super-resolution reconstruction (SRR) for HR visualization in the standard anatomical space. We validate our framework with images from fetuses with variable degrees of ventriculomegaly associated with spina bifida. Experiments show that each step of our proposed pipeline outperforms state-of-the-art methods in both segmentation and reconstruction comparisons. Overall, we report automatic SRR reconstructions that compare favorably with those obtained by manual, labor-intensive brain segmentations. This potentially unlocks the use of automatic fetal brain reconstruction studies in clinical practice.

Arachnoid Cyst—Prenatal Detection, Management, and Perinatal Outcome

Abstract We describe the natural history, associated abnormalities, and outcome of fetuses with arachnoid cyst (AC) diagnosed antenatally by ultrasound and magnetic resonance imaging (MRI). Most congenital ACs are diagnosed from mid-trimester onwards. They are mostly well-delineated and unilocular. The supratentorial ACs are picked up later than those in the posterior fossa. Fetal MRI improves the visualization of associated central nervous system abnormalities. The majority of ACs are of benign origin and remain stable. Based on the current series and the review of the literature, in the absence of other associated anomalies and when the karyotype is normal, the postnatal overall and neurological outcome is favorable. In the postnatal workup special attention should be paid to symptoms of increased intracranial pressure. The large majority do not require surgery. Slow growth of the cyst, absence of ventriculomegaly, and the absence of associated central nervous anomalies are favorable variables that can be used in the prenatal counseling.

P02.07: Termination of pregnancy for midline cerebral anomalies

Objectives: To compare the accuracy of neurosonography (NSG) and magnetic resonance imaging (MRI) in the assessment of the fetal cortex within a high risk cohort of fetuses diagnosed with central nervous system (CNS) abnormalities. Methods: Single centre prospective study involving an Italian Tertiary Unit. The included cases had antenatal diagnosis or suspicion of CNS abnormality and underwent twoand three-dimensional NSG, which was performed using a 3-5 MHz and a 4-6 MHz endocavitary probe equipped Samsung WS-80 scanner, and MRI. Antenatal findings were compared between the two techniques and differences in terms of CNS anatomy and cortical findings were recorded. Postnatal imaging was also investigated for the study purposes. Results: In all, 31 cases were submitted to NSG at a median gestation of 26+0 weeks (18+3 – 33+2) due to one of the following: abnormal or absent cavum septum pellucidum (5 cases), ventriculomegaly (9 cases), head circumference <3SD (4 cases), posterior fossa abnormalities (3 cases) or other reasons (8 cases), including previous intracranial anomaly or fetal CMV infection. At expert NSG, abnormal cortical findings were diagnosed in 8/31 cases, including lissencephaly (3 cases), microcephaly (1 case) and focal cortical dysplasia in the remaining. All fetuses underwent antenatal MRI, which was performed at a median gestation of 28+4 weeks (20+0 – 34+2). MRI yielded abnormal cortical findings consistent with those reported at NSG in all 8 cases. Additionally, mild asymmetry in the cortical sulci and gyri was noted at MRI in 2 of the remaining 23 cases with normal cortical appearance at NSG. Antenatal findings were confirmed at postnatal imaging or post-mortem examination in all cases. Conclusions: Our results from a cohort at high risk of CNS abnormalities have suggested that antenatal MRI is not superior to expert NSG in the detection of abnormalities of the cortical folding and development.

OP04.02: Congenital diaphragmatic hernia and diaphragmatic eventration: separate entities or two sides of the same coin?

N. Abbasi1, R. Ruano2, A. Johnson3, S. Haleh4, A. Benachi5, J. Saada5, G. Ryan1, FETO Consortium 1Mount Sinai Hospital, University of Toronto, Toronto, ON, Canada; 2Department of Obstetrics and Gynecology, Mayo Clinic, Rochester, MN, USA; 3Obstetrics and Gynecology, UTHSC, Houston, TX, USA; 4Obstetrics and Gynecology, Texas Children’s Hospital, Baylor College of Medicine, Houston, TX, USA; 5Hôpital Antoine-Béclère, Hôpitaux Universitaires Paris-Sud, Paris, France