Dissertation•
Minimally interactive segmentation with application to human placenta in fetal MR images
28 Jun 2018-
TL;DR: Experimental results show that the proposed algorithms outperform traditional interactive segmentation methods in terms of accuracy and interactivity, and might be suitable for segmentation of the placenta in planning systems for fetal and maternal surgery, and for rapid characterization of theplacenta by MR images.
Abstract: Placenta segmentation from fetal Magnetic Resonance (MR) images is important for fetal surgical planning. However, accurate segmentation results are difficult to achieve for automatic methods, due to sparse acquisition, inter-slice motion, and the widely varying position and shape of the placenta among pregnant women. Interactive methods have been widely used to get more accurate and robust results. A good interactive segmentation method should achieve high accuracy, minimize user interactions with low variability among users, and be computationally fast. Exploiting recent advances in machine learning, I explore a family of new interactive methods for placenta segmentation from fetal MR images. I investigate the combination of user interactions with learning from a single image or a large set of images. For learning from a single image, I propose novel Online Random Forests to efficiently leverage user interactions for the segmentation of 2D and 3D fetal MR images. I also investigate co-segmentation of multiple volumes of the same patient with 4D Graph Cuts. For learning from a large set of images, I first propose a deep learning-based framework that combines user interactions with Convolutional Neural Networks (CNN) based on geodesic distance transforms to achieve accurate segmentation and good interactivity. I then propose image-specific fine-tuning to make CNNs adaptive to different individual images and able to segment previously unseen objects. Experimental results show that the proposed algorithms outperform traditional interactive segmentation methods in terms of accuracy and interactivity. Therefore, they might be suitable for segmentation of the placenta in planning systems for fetal and maternal surgery, and for rapid characterization of the placenta by MR images. I also demonstrate that they can be applied to the segmentation of other organs from 2D and 3D images.
Citations
More filters
Journal Article•
TL;DR: In this paper, the authors provide an overview of the rapidly developing field of photoacoustic imaging, which is a promising method for visualizing biological tissues with optical absorbers, compared with optical imaging and ultrasonic imaging.
Abstract: Photoacoustic imaging is a promising method for visualizing biological tissues with optical absorbers. This article provides an overview of the rapidly developing field of photoacoustic imaging. Photoacoustics, the physical basis of photoacoustic imaging, is analyzed briefly. The merits of photoacoustic technology, compared with optical imaging and ultrasonic imaging, are described. Various imaging techniques are also discussed, including scanning tomography, computed tomography and original detection of photoacoustic imaging. Finally, some biomedical applications of photoacoustic imaging are summarized.
618 citations
Journal Article•
TL;DR: Endoscopic laser coagulation of anastomoses is a more effective first linetreatment than serial amnioreduction for severe twin totwin transfusion syndrome diagnosed before 26 weeks of gestation.
Abstract: BACKGROUND: Monochorionic twin pregnancies complicated by severe twin to twin transfusion syndrome at midgestation can be treated by either serial amnioreduction(removal of large volumes of amniotic fluid) or selective fetoscopic laser coagulation of the communicating vessels on the chorionic plate. We conducted a randomized trial to compare the efficacy and safety of these two treatments. METHODS: Pregnant women with severe twin to twin transfusion syndrome before 26 weeks of gestation were randomly assigned to laser therapy or amnioreduction. We assessed perinatal survival of at least one twin (a prespecified primary outcome),survival of at least one twin at six months of age, and survival without neurologic complications at six months of age on the basis of the number of pregnancies or the number of fetuses or infants, as appropriate. RESULTS:The study was concluded early, after 72 women had been assigned to the laser group and 70 to the amnioreduction group, because a planned interim analysis demonstrated a significant benefit in the laser group. As compared with the amnioreduction group, the laser group had a higher likelihood of the survival of at least one twin to 28 days of age (76 percent vs. 56 percent; relative risk of the death of both fetuses, 0.63; 95 percent confidence interval, 0.25to 0.93; P=0.009) and 6 months of age (P=0.002). Infants in the laser group also had a lower incidence of cystic periventricular leukomalacia (6 percent vs. 14 percent,P=0.02) and were more likely to be free of neurologic complications at six months of age (52 percent vs. 31 percent, P=0.003). CONCLUSIONS: Endoscopic laser coagulation of anastomoses is a more effective first linetreatment than serial amnioreduction for severe twin totwin transfusion syndrome diagnosed before 26 weeks of gestation.
207 citations
10 Jan 2015
TL;DR: In this paper, the authors proposed an early diagnosis of chorionicity, amnionicity and identification of placental anomalies for the adequate management of twin pregnancies, which can help in assessing the presence of placenta and umbilical cord abnormalities.
Abstract: The frequency of twin gestations has increased over the last few decades, mainly due to maternal age at childbearing, and the use of assisted reproductive technologies. Twins are at higher risk of aneuploidy, structural anomalies, and placental abnormalities. Some of the placental and umbilical cord abnormalities found in twin gestations are nonspecific and can be found in singleton gestations (ie, placenta previa, placental abruption, single umbilical artery, velamentous cord insertion, vasa previa, etc). However, other anomalies are unique to twin gestations, and are mainly associated with monochorionic twins-these include intraplacental anastomosis and cord entanglement. Most of these conditions can be diagnosed with ultrasound. An accurate and early diagnosis is important in the management of twin gestations. Determination of chorionicity, amnionicity, and the identification of placental anomalies are key issues for the adequate management of twin pregnancies. Pathologic placental examination after delivery can help in assessing the presence of placental and umbilical cord abnormalities, as well as providing information about chorionicity and gaining insight into the potential mechanisms of disease affecting twin gestations.
21 citations
References
More filters
27 Jun 2016
TL;DR: In this article, the authors proposed a residual learning framework to ease the training of networks that are substantially deeper than those used previously, which won the 1st place on the ILSVRC 2015 classification task.
Abstract: Deeper neural networks are more difficult to train. We present a residual learning framework to ease the training of networks that are substantially deeper than those used previously. We explicitly reformulate the layers as learning residual functions with reference to the layer inputs, instead of learning unreferenced functions. We provide comprehensive empirical evidence showing that these residual networks are easier to optimize, and can gain accuracy from considerably increased depth. On the ImageNet dataset we evaluate residual nets with a depth of up to 152 layers—8× deeper than VGG nets [40] but still having lower complexity. An ensemble of these residual nets achieves 3.57% error on the ImageNet test set. This result won the 1st place on the ILSVRC 2015 classification task. We also present analysis on CIFAR-10 with 100 and 1000 layers. The depth of representations is of central importance for many visual recognition tasks. Solely due to our extremely deep representations, we obtain a 28% relative improvement on the COCO object detection dataset. Deep residual nets are foundations of our submissions to ILSVRC & COCO 2015 competitions1, where we also won the 1st places on the tasks of ImageNet detection, ImageNet localization, COCO detection, and COCO segmentation.
123,388 citations
Proceedings Article•
01 Jan 2015TL;DR: This work introduces Adam, an algorithm for first-order gradient-based optimization of stochastic objective functions, based on adaptive estimates of lower-order moments, and provides a regret bound on the convergence rate that is comparable to the best known results under the online convex optimization framework.
Abstract: We introduce Adam, an algorithm for first-order gradient-based optimization of stochastic objective functions, based on adaptive estimates of lower-order moments. The method is straightforward to implement, is computationally efficient, has little memory requirements, is invariant to diagonal rescaling of the gradients, and is well suited for problems that are large in terms of data and/or parameters. The method is also appropriate for non-stationary objectives and problems with very noisy and/or sparse gradients. The hyper-parameters have intuitive interpretations and typically require little tuning. Some connections to related algorithms, on which Adam was inspired, are discussed. We also analyze the theoretical convergence properties of the algorithm and provide a regret bound on the convergence rate that is comparable to the best known results under the online convex optimization framework. Empirical results demonstrate that Adam works well in practice and compares favorably to other stochastic optimization methods. Finally, we discuss AdaMax, a variant of Adam based on the infinity norm.
111,197 citations
"Minimally interactive segmentation ..." refers methods in this paper
...g, using adaptive sub-gradient (AdaGrad) [183], adaptive learning rate (RMSProp) [184], and adaptive moment estimation (Adam) [185]....
[...]
Proceedings Article•
03 Dec 2012TL;DR: The state-of-the-art performance of CNNs was achieved by Deep Convolutional Neural Networks (DCNNs) as discussed by the authors, which consists of five convolutional layers, some of which are followed by max-pooling layers, and three fully-connected layers with a final 1000-way softmax.
Abstract: We trained a large, deep convolutional neural network to classify the 1.2 million high-resolution images in the ImageNet LSVRC-2010 contest into the 1000 different classes. On the test data, we achieved top-1 and top-5 error rates of 37.5% and 17.0% which is considerably better than the previous state-of-the-art. The neural network, which has 60 million parameters and 650,000 neurons, consists of five convolutional layers, some of which are followed by max-pooling layers, and three fully-connected layers with a final 1000-way softmax. To make training faster, we used non-saturating neurons and a very efficient GPU implementation of the convolution operation. To reduce overriding in the fully-connected layers we employed a recently-developed regularization method called "dropout" that proved to be very effective. We also entered a variant of this model in the ILSVRC-2012 competition and achieved a winning top-5 test error rate of 15.3%, compared to 26.2% achieved by the second-best entry.
73,978 citations
Proceedings Article•
01 Jan 2015TL;DR: In this paper, the authors investigated the effect of the convolutional network depth on its accuracy in the large-scale image recognition setting and showed that a significant improvement on the prior-art configurations can be achieved by pushing the depth to 16-19 layers.
Abstract: In this work we investigate the effect of the convolutional network depth on its accuracy in the large-scale image recognition setting. Our main contribution is a thorough evaluation of networks of increasing depth using an architecture with very small (3x3) convolution filters, which shows that a significant improvement on the prior-art configurations can be achieved by pushing the depth to 16-19 weight layers. These findings were the basis of our ImageNet Challenge 2014 submission, where our team secured the first and the second places in the localisation and classification tracks respectively. We also show that our representations generalise well to other datasets, where they achieve state-of-the-art results. We have made our two best-performing ConvNet models publicly available to facilitate further research on the use of deep visual representations in computer vision.
49,914 citations
05 Oct 2015
TL;DR: Neber et al. as discussed by the authors proposed a network and training strategy that relies on the strong use of data augmentation to use the available annotated samples more efficiently, which can be trained end-to-end from very few images and outperforms the prior best method (a sliding-window convolutional network) on the ISBI challenge for segmentation of neuronal structures in electron microscopic stacks.
Abstract: There is large consent that successful training of deep networks requires many thousand annotated training samples. In this paper, we present a network and training strategy that relies on the strong use of data augmentation to use the available annotated samples more efficiently. The architecture consists of a contracting path to capture context and a symmetric expanding path that enables precise localization. We show that such a network can be trained end-to-end from very few images and outperforms the prior best method (a sliding-window convolutional network) on the ISBI challenge for segmentation of neuronal structures in electron microscopic stacks. Using the same network trained on transmitted light microscopy images (phase contrast and DIC) we won the ISBI cell tracking challenge 2015 in these categories by a large margin. Moreover, the network is fast. Segmentation of a 512x512 image takes less than a second on a recent GPU. The full implementation (based on Caffe) and the trained networks are available at http://lmb.informatik.uni-freiburg.de/people/ronneber/u-net .
49,590 citations