Browsing by Author "Merhof, Dorit"

Now showing 1 - 5 of 5
  • Thumbnail Image
    Item
    Annotated Dendrograms for Neurons From the Larval Fruit Fly Brain
    (The Eurographics Association, 2018) Strauch, Martin; Hartenstein, Volker; Andrade, Ingrid V.; Cardona, Albert; Merhof, Dorit; Puig Puig, Anna and Schultz, Thomas and Vilanova, Anna and Hotz, Ingrid and Kozlikova, Barbora and Vázquez, Pere-Pau
    Recent advances in neuroscience have made it possible to reconstruct all neurons of an entire neural circuit in the larval fruit fly brain from serial electron microscopy image stacks. The reconstructed neurons are morphologically complex 3D graphs whose nodes are annotated with labels representing different types of synapses. Here, we propose a method to draw simplified, yet realistic 2D neuron sketches of insect neurons in order to help biologists formulate hypotheses on neural function at the microcircuit level. The sketches are dendrograms that capture a neuron's branching structure and that preserve branch lengths, providing realistic estimates for distances and signal travel times between synapses. To improve readability of the often densely clustered synapse annotations, synapses are automatically summarized in local clusters of synapses of the same type and arranged to minimize label overlap. We show that two major neuron classes of an olfactory circuit in the larval fruit fly brain can be discriminated visually based on the dendrograms. Unsupervised and supervised data analysis reveals that class discrimination can be performed using morphological features derived from the dendrograms.
  • Thumbnail Image
    Item
    Eurographics Workshop on Visual Computing for Biology and Medicine 2017: Frontmatter
    (Eurographics Association, 2017) Bruckner, Stefan; Hennemuth, Anja; Kainz, Bernhard; Hotz, Ingrid; Merhof, Dorit; Rieder, Christian; Stefan Bruckner and Anja Hennemuth and Bernhard Kainz and Ingrid Hotz and Dorit Merhof and Christian Rieder
  • Thumbnail Image
    Item
    Introducing CNN-Based Mouse Grim Scale Analysis for Fully Automated Image-Based Assessment of Distress in Laboratory Mice
    (The Eurographics Association, 2018) Kopaczka, Marcin; Ernst, Lisa; Schock, Justus; Schneuing, Arne; Guth, Alexander; Tolba, Rene; Merhof, Dorit; Puig Puig, Anna and Schultz, Thomas and Vilanova, Anna and Hotz, Ingrid and Kozlikova, Barbora and Vázquez, Pere-Pau
    International standards require close monitoring of distress of animals undergoing laboratory experiments in order to minimize the stress level and allow choosing minimally stressful procedures for the experiments. Currently, one of the the best established severity assessment procedures is the mouse grimace scale (MGS), a protocol in which images of the animals are taken and scored by assessing five key visual features that have been shown to be highly correlated with distress and pain. While proven to be highly reliable, MGS assessment is currently a time-consuming task requiring manual video processing for key frame extraction and subsequent expert grading. Additionally, due to the the high per-picture expert time required, MGS scoring is performed on a small number of selected frames from a video. To address these shortcomings, we introduce a method for fully automated real-time MGS scoring of orbital eye tightening, one of the five sub-scores. We define and evaluate the method which is centered around a set of convolutional neural networks (CNNs) and allows live continuous MGS assessment of a mouse in real time. We additionally describe a multithreaded client-server architecture with a graphical user interface that allows convenient use of the developed method for simultaneous real-time MGS scoring of several animals.
  • Thumbnail Image
    Item
    Multiparametric Magnetic Resonance Image Synthesis using Generative Adversarial Networks
    (The Eurographics Association, 2019) Haarburger, Christoph; Horst, Nicolas; Truhn, Daniel; Broeckmann, Mirjam; Schrading, Simone; Kuhl, Christiane; Merhof, Dorit; Kozlíková, Barbora and Linsen, Lars and Vázquez, Pere-Pau and Lawonn, Kai and Raidou, Renata Georgia
    Generative adversarial networks have been shown to alleviate the problem of limited training data for supervised learning problems in medical image computing. However, most generative models for medical images focus on image-to-image translation rather than de novo image synthesis. In many clinical applications, image acquisition is multiparametric, i.e. includes contrast-enchanced or diffusion-weighted imaging. We present a generative adversarial network that synthesizes a sequence of temporally consistent contrast-enhanced breast MR image patches. Performance is evaluated quantitatively using the Fréchet Inception Distance, achieving a minimum FID of 21.03. Moreover, a qualitative human reader test shows that even a radiologist cannot differentiate between real and fake images easily.
  • Thumbnail Image
    Item
    Semantic Segmentation of Brain Tumors in MRI Data Without any Labels
    (The Eurographics Association, 2019) Weninger, Leon; Krauhausen, Imke; Merhof, Dorit; Kozlíková, Barbora and Linsen, Lars and Vázquez, Pere-Pau and Lawonn, Kai and Raidou, Renata Georgia
    Brain MR images are one of the most important instruments for diagnosing neurological disorders such as tumors, infections or trauma. In particular, grade I-IV brain tumors are a well-studied subject for supervised deep learning approaches. However, for a clinical use of these approaches, a very large annotated database that covers all of the occurring variance is necessary. As MR scanners are not quantitative, it is unclear how good supervised approaches, trained on a specific database, will actually perform on a new set of images that may stem from a yet other scanner. We propose a new method for brain tumor segmentation, that can not only identify abnormal regions, but can also delineate brain tumors into three characteristic radiological areas: The edema, the enhancing core, and the non-enhancing and necrotic tissue. Our concept is based on FLAIR and T1CE MRI sequences, where abnormalities are detected with a variational autoencoder trained on healthy examples. The detected areas are finally postprocessed via Gaussian Mixture Models and finally classified according to the three defined labels. We show results on the BraTS2018 dataset and compare these to previously published unsupervised segmentation results as well as to the results of the BraTS challenge 2018. Our developed unsupervised anomaly detection approach is on par with previously published methods. Meanwhile, the semantic segmentation - a new and unique model - shows encouraging results.