Stefan Zachow

Bio: Stefan Zachow is an academic researcher from Zuse Institute Berlin. The author has contributed to research in topics: Segmentation & Active shape model. The author has an hindex of 22, co-authored 37 publications receiving 1484 citations.

Fusion of computed tomography data and optical 3D images of the dentition for streak artefact correction in the simulation of orthognathic surgery

Abstract: Objective: To determine the limits of accuracy of fusion of optical three-dimensional (3D) imaging and computed tomography (CT) with and without metal artefacts in an experimental setting and to show the application of this hybrid system in 3D orthognathic surgery simulation. Methods: Ten plaster casts of dental arches were subjected to a CT scan and optical 3D surface imaging. Subsequently, the first molars in the plaster casts were supplied with metal restorations, bilaterally, and new CT scans and optical surface images were assessed. The registration of the surface data of the two imaging modalities of the study models without and with metal restorations was carried out. The mean distance between the two data sets was calculated. From a patient a CT scan of the skull as well as optical 3D images of plaster casts of the dental arches were acquired. Again the two imaging modalities were registered and virtual orthognathic surgery simulation was carried out. Results: The mean distance between the corresp...

Automatic segmentation of the pelvic bones from CT data based on a statistical shape model

Abstract: We present an algorithm for automatic segmentation of the human pelvic bones from CT datasets that is based on the application of a statistical shape model. The proposed method is divided into three steps: 1) The averaged shape of the pelvis model is initially placed within the CT data using the Generalized Hough Transform, 2) the statistical shape model is then adapted to the image data by a transformation and variation of its shape modes, and 3) a final free-form deformation step based on optimal graph searching is applied to overcome the restrictive character of the statistical shape representation. We thoroughly evaluated the method on 50 manually segmented CT datasets by performing a leave-one-out study. The Generalized Hough Transform proved to be a reliable method for an automatic initial placement of the shape model within the CT data. Compared to the manual gold standard segmentations, our automatic segmentation approach produced an average surface distance of 1.2 ± 0.3mm after the adaptation of the statistical shape model, which could be reduced to 0.7±0.3mm using a final free-form deformation step. Together with an average segmentation time of less than 5 minutes, the results of our study indicate that our method meets the requirements of clinical routine.

Three-dimensional osteotomy planning in maxillofacial surgery including soft tissue prediction.

Abstract: Preoperative planning of complex osteotomies in craniomaxillofacial surgery, in conjunction with a surgeon's expertise, is essential for achieving an optimal result. However, the soft tissue changes that accompany facial bone movements cannot yet be accurately predicted. Bony tissue, because of its

The digital bee brain: integrating and managing neurons in a common 3D reference system

Abstract: The honeybee standard brain (HSB) serves as an interactive tool for relating morphologies of bee brain neurons and provides a reference system for functional and bibliographical properties (http://www.neurobiologie.fu-berlin.de/beebrain/). The ultimate goal is to document not only the morphological network properties of neurons collected from separate brains, but also to establish a graphical user interface for a neuron-related data base. Here, we review the current methods and protocols used to incorporate neuronal reconstructions into the HSB. Our registration protocol consists of two separate steps applied to imaging data from two-channel confocal microscopy scans: (1) The reconstruction of the neuron, facilitated by an automatic extraction of the neuron’s skeleton based on threshold segmentation, and (2) the semi-automatic 3D segmentation of the neuropils and their registration with the HSB. The integration of neurons in the HSB is performed by applying the transformation computed in step (2) to the reconstructed neurons of step (1). The most critical issue of this protocol in terms of user interaction time – the segmentation process – is drastically improved by the use of a model-based segmentation process. Furthermore, the underlying statistical shape models (SSM) allow the visualization and analysis of characteristic variations in large sets of bee brain data. The anatomy of neural networks composed of multiple neurons that are registered into the HSB are visualized by depicting the 3D reconstructions together with semantic information with the objective to integrate data from multiple sources (electrophysiology, imaging, immunocytochemistry, molecular biology). Ultimately, this will allow the user to specify cell types and retrieve their morphologies along with physiological characterizations.

The Liver Tumor Segmentation Benchmark (LiTS)

Abstract: In this work, we report the set-up and results of the Liver Tumor Segmentation Benchmark (LITS) organized in conjunction with the IEEE International Symposium on Biomedical Imaging (ISBI) 2016 and International Conference On Medical Image Computing Computer Assisted Intervention (MICCAI) 2017. Twenty four valid state-of-the-art liver and liver tumor segmentation algorithms were applied to a set of 131 computed tomography (CT) volumes with different types of tumor contrast levels (hyper-/hypo-intense), abnormalities in tissues (metastasectomie) size and varying amount of lesions. The submitted algorithms have been tested on 70 undisclosed volumes. The dataset is created in collaboration with seven hospitals and research institutions and manually reviewed by independent three radiologists. We found that not a single algorithm performed best for liver and tumors. The best liver segmentation algorithm achieved a Dice score of 0.96(MICCAI) whereas for tumor segmentation the best algorithm evaluated at 0.67(ISBI) and 0.70(MICCAI). The LITS image data and manual annotations continue to be publicly available through an online evaluation system as an ongoing benchmarking resource.

The honeybee as a model for understanding the basis of cognition.

Abstract: Honeybees contradict the notion that insect behaviour tends to be relatively inflexible and stereotypical. Indeed, they live in colonies and exhibit complex social, navigational and communication behaviours, as well as a relatively rich cognitive repertoire. Because these relatively complex behaviours are controlled by a brain consisting of only 1 million or so neurons, honeybees offer an opportunity to study the relationship between behaviour and cognition in neural networks that are limited in size and complexity. Most recently, the honeybee has been used to model learning and memory formation, highlighting its utility for neuroscience research, in particular for understanding the basis of cognition.

Invertebrate learning and memory: Fifty years of olfactory conditioning of the proboscis extension response in honeybees

Abstract: The honeybee Apis mellifera has emerged as a robust and influential model for the study of classical conditioning, thanks to the existence of a powerful Pavlovian conditioning protocol, the olfactory conditioning of the proboscis extension response (PER) In 2011, the olfactory PER conditioning protocol celebrates 50 years since it was first introduced by Kimihisa Takeda in 1961 Here, we review its origins, developments, and perspectives in order to define future research avenues and necessary methodological and conceptual evolutions We show that olfactory PER conditioning has become a versatile tool for the study of questions in extremely diverse fields in addition to the study of learning and memory and that it has allowed behavioral characterizations, not only of honeybees, but also of other insect species, for which the protocol was adapted We celebrate, therefore, Takeda’s original work and prompt colleagues to conceive and establish further robust behavioral tools for an accurate characterization of insect learning and memory at multiple levels of analysis