The sponsors and projects for 2017 include:
Within MedLab – and especially the Therapy Planning group at ZIB – we are dealing with a variety of medical and anthropometric data. To tackle the challenges of analyzing an always increasing amount of data and to provide software tools to automatically extract the relevant information out of it, we are investigating model based approaches (statistical shape and appearance models) as well as machine learning techniques (regression, classification, and semi-supervised learning), which can then be used in a number of applications such as scene recognition from photographs, object recognition in images, parameter estimation, classification, up to automatic diagnosis from medical image data.
The project is in close collaboration with 1000shapes GmbH, a ZIB spin-off that transfers research in life sciences into products for clinical applications. Within the project, algorithms are to be identified, developed/implemented, compared, and tested on data from various application domains. The successful applicant will have the opportunity to perform research in medical image computing and geometry processing within the ZIB research group Therapy Planning while obtaining professional support from 1000shapes in software development with insights into relevant applications. Within the project, students will have the opportunity to experience medical research under professional supervision in combination with industry-strength software development, with the goal of practical solutions and publication.
Building on large medical image as well as a anthropometric 3D face databases, students will have the opportunity to investigate machine-learning approaches, i.e. deep learning, or the application of regression forests, to identify, analyze, and classify features or patterns based on medical images or geometric models.
Several databases are the foundation for possible investigations: (1) The OAI database of the Osteoarthritis Initiative (OAI), which is a multi-center, longitudinal, prospective observational study of knee osteoarthritis, providing clinical evaluation data, radiological (x-ray and magnetic resonance) images, and a bio-specimen repository from over 5000 patients. This information has great potential, both for developing a better understanding of disease onset and progression, as well as improving future therapeutic concepts; (2) A database of several hundreds of 3D face models from various individuals and with varying facial expressions, providing information for anthropometrical studies or psychological experiments; (3) A huge collection of dental 3D image data, where bony structures, nerves, and teeth are to be segmented, anatomical relationships are to be analyzed, and suitable shape and appearance models are to be developed; (4) A database of human spines giving the opportunity to study the morphology of single vertebrae up to the complete spine as well as the functional performance within the context of biomechanics and orthopedic research. The processing of such databases requires automated image and geometry processing as well as sophisticated data analysis approaches.
For medical image processing, the challenge is to automatically extract anatomical shape and appearance information from image data, as well as to integrate this information in so called statistical 3D shape and appearance models to train and improve automated algorithms. For geometric data the challenge is to improve methods for determining correspondences, to analyze variation in shape, to establish suitable metrics for measuring similarities in various shape spaces, for clustering and population based analysis. To this end, machine learning combined with model-based approaches shall be employed and adopted.
Establishing correspondences between shapes lies at the core of many operations in image analysis and geometry processing. The majority of existing methods formulate the matching problem as finding optimal pairings of points or regions on shapes. This representation, however, renders the matching intractable as the space of possible point correspondences grows exponentially and does not naturally support constraints such as map continuity or global consistency. Within a possible project, we will investigate a recent alternative approach that generalizes the notion of correspondences to mappings between real-valued functions on the shapes rather than the standard point-to-point maps. One challenge that we will address is the adaptation of this functional maps framework to the matching of volumetric geometries from 3D medical image data. Based on this, we will further derive a scheme for inferring group-wise correspondences that takes advantage of the context provided by the collection of shapes.
Based on the aforementioned background several topics in medical image and geometry processing are conceivable. Results of our research may become a basis for improving existing segmentation, classification, and diagnosis algorithms that are currently under development by 1000shapes.
The prospective participant should:
The computational nanooptics group at ZIB investigates advanced numerical techniques for simulating the interaction of light and nanoscale objects. The numerical methods developed and investigated in the group include finite-element methods with h-, p-, and hp-adaptivity, reduced basis methods, discontinuous Galerkin methods, and others. Applications range from fundamental research in physics to device design in the optical and semiconductor industries. This includes topics like optimization of photovoltaic devices for improved conversion efficiency, optimization of nanophotonic devices for quantum optics applications like quantum cryptography, design of plasmonic nano-antennas for optical near-field sensing, computational lithography for the design of state-of-the-art photolithography masks for computer chip manufacturers, 3D metamaterial design, and other topics.
JCMwave GmbH is a ZIB spin-off which develops and provides state-of-the-art finite element software. Within the JCMwave infrastructures the students of this project will have the opportunity to work with the newest development versions of finite-element software, to discuss with the development team in regular meetings, and to get an insight to industrial nanotechnology design challenges.
You will learn how to model and simulate nanophotonic setups. The underlying physical model is typically Maxwell’s wave equation in three spatial dimensions. A main challenge in such simulations is to obtain simulation results with upper bounds to numerical discretization errors within short computation times. Accurate and fast results are required e.g. for design optimizations in high-dimensional parameter spaces, and for parameter retrieval in optical metrology. For in-line applications in industrial quality control, speed and accuracy of parameter retrieval is currently a limiting factor to production speed. As shown in various benchmarks, the finite-element method is well suited to handle such computations, as its performance for highly accurate results can be orders of magnitude faster than competing methods. However, to further improve on its performance various properties of the method and of the models of interest can be exploited. These include higher-order vectorial finite elements, adaptive mesh refinement, hp-adaptivity, and automatic differentiation. This project will also concentrate on recent developments exploiting symmetries of the underlying models.
It is planned that the team members will specialize in the fields of mesh generation, finite element convergence and post-processing techniques, respectively. The team will then join the experiences from these fields to investigate methods for improved simulation efficiency, exploiting symmetries of modern nanophotonic devices. This includes validating results from automatic symmetry-detection methods.
The project should result in a comprehensive report to be presented at the end, and in a collection of Matlab- or Python-based automatic test routines. We expect that the report will meet high standards as we aim at a joint publication of the results in a peer-reviewed journal.
We welcome applications from highly motivated team-players who ideally
A major aim of the Research Campus MODAL is the development and use of mathematical synergies between the individual labs of the network. In this context, MODAL SynLab, the Synergy Laboratory, aims to generalize problem-specific solution algorithms developed in each laboratory and implement it in a structure-specific way. At the core of these activities is the development of the SCIP Optimization Suite, a software package for solving constraint and mixed-integer nonlinear optimization problems to global optimality. Learning from the problem-specific successes in the application-specific labs, our main goal is to accelerate the existing solution algorithms and extend the class of problems that can be handled. This way we create an optimization tool that can form a stronger basis for future research projects.
Satalia is an optimization solutions company. Based in London, they develop SolveEngine, a platform that aims to make optimization technologies more accessible to practitioners. The company also produces stand-alone optimization tools, which it uses in some of its consulting projects for customers worldwide. Across the company, they use a diverse set of exact and heuristic algorithms such as satisfiability solving, mixed-integer linear programming, and constraint programming, both in general and problem-specific implementations. It is their declared mission to transfer cutting-edge optimization technology developed in academia to practice. The G-RIPS project will be accompanied by their Berlin-based consultants.
Satalia’s SolveEngine interfaces to a variety of optimization algorithms. A given optimization problem can often be solved by many of these algorithms, but their performance can vary widely in practice. Predicting the best-performing solver on-the-fly and under limited response time is an unsolved question. The aim of this project is to investigate and compare different machine learning techniques in order to select well-performing algorithms from instance features that can be collected with limited computational effort. This will involve understanding various machine learning algorithms, the design of experiments, and the use of existing machine learning packages and own implementations. Practitioners from Satalia and mixed-integer programming solver developers from MODAL SynLab will provide support. Students will gain experience in the implementation and use of machine learning as a research tool and gain insight into the world of mixed-integer programming algorithms.
We welcome applications from highly motivated team-players who ideally