Deformable Registration of Medical Images Using Soft Computing (European Centre for Soft Computing, Spain)
The overall aim of the subproject is to investigate the use of Soft Computing to tackle the deformable registration of medical images following a two-stage approach. First, a rigid and affine registration process is tackled using bio-inspired computing and different voxel similarity measures. Then, a deformable IR method aims to refine the results from the previous stage.
Genetic Fuzzy Systems and Deformable Models for Medical Image Segmentation (European Centre for Soft Computing, Spain)
This subproject aims at investigating the use of fuzzy systems and deformable models for the segmentation of objects of interest in medical images. Besides, following an evolutionary learning process, we aim to customize the latter systems to allow them to be applied to a wide branch of medical imaging scenarios.
Image Restoration: Noise Reduction (University of Gent, Belgium)
The overall aim of the subproject is to investigate how soft computing techniques, and in particular fuzzy techniques, can contribute to the development of high performance noise reduction filters for medical images. Noise reduction is important to improve the quality of an image for visualization (preserving diagnostic details) or as a pre-processing step before applying other processing tasks such as edge detection, segmentation, or compression.
Image Restoration: Similarity Measures (University of Gent, Belgium)
This subproject aims at investigating the construction of objective quality measures or measures of comparison, specifically for medical images, using soft computing techniques in general and fuzzy techniques in particular. Such measures are important in the context of image retrieval (retrieval of medical images from a database that are most similar to a reference image can support diagnosis), and also allow to automatically evaluate image quality.
Application of Automatic Algorithm Configuration in Image Analysis (Université Libre de Bruxelles, Belgium)
In this project the application and improvement of tools for the automatic configuration of algorithms will be investigated. In fact, a challenging and cumbersome task in the application of many algorithmic techniques is to determine appropriate values for free parameters and only few approaches have been proposed for automating this task. In particular, studies will be focused on the development of automatic configuration techniques specifically tailored for image analysis problems and apply the configuration algorithms developed in complex algorithm configuration tasks such as they arise, for example, in various applications considered in MIBISOC and in the image analysis domain more in general.
Development of Tools for Automatic Algorithm Configuration (Université Libre de Bruxelles, Belgium)
The overall aim of the subproject is to investigate methods for the automatic configuration of algorithms. In fact, a challenging and cumbersome task in the application of many algorithmic techniques is to determine appropriate values for free parameters and only few approaches have been proposed for automating this task. In particular, we will (i) experimentally assess currently available algorithm configuration tools; (ii) further develop the set of available configuration algorithms; and (iii) study possible ways of interaction between off-line algorithm configuration and techniques that allow the online adaptation of parameters while actually solving a problem.
Manifold Learning for Medical Imaging: developing a 3D interactive MR image segmentation system (University of Nottingham, UK)
This subproject aims at investigating the use of Mathematical, Soft Computing and Machine Learning techniques for processing analysing medical images, in particular, Magnetic Resonance Images, and to use the analysis to help develop biomarkers for diseases. Specifically, the objectives of this subproject will be to develop a 3D interactive MR image segmentation system. Knowledge in deformable object modelling, variational, and statistical methods will be an advantage.
Manifold Learning for Medical Imaging: developing methods for identifying non-linear structures in medical imaging data (University of Nottingham, UK)
The overall aim of the subproject is to investigate the use of Mathematical, Soft Computing and Machine Learning techniques for processing analysing medical images, in particular, Magnetic Resonance Images, and use the analysis to help develop biomarkers for diseases. Specifically, the subproject objectives will be to develop methods for identifying non-linear structures in medical imaging data. Knowledge in manifold learning and soft computing will an advantage.
Detection and analysis of anatomical structures in multi-dimensional image sets (Universitá degli Studi di Parma, Italy)
This subproject is focused on the investigation of techniques for detection of anatomical districts of interest, to be used in tasks like: 3D reconstruction and visualization, detection of morphological or structural anomalies, matching of real images with anatomical atlases, extraction of features of clinical/physiological interest, etc.
Bio-inspired techniques for multi-dimensional image analysis (Universitá degli Studi di Parma, Italy)
The overall aim of the subproject is to investigate novel approaches to image analysis fully based on bio-inspired algorithms, or in which bio-inspired algorithms are an essential part. Of particular interest for the project will be the analysis of multi-dimensional data sets, such as image sequences, 3D sets of tomographic images, etc.
Multi-objective Genetic Fuzzy Systems (University of Granada, Spain)
The subproject will investigate the development of new approaches for learning fuzzy models by means of multi-objective genetic algorithms and their use for classification, imbalanced data sets, etc.
Evolutionary Algorithms for continuous optimization: Large scale optimization and Applications (University of Granada, Spain)
The development of new approaches for parameter optimization with evolutionary algorithms, in particular, by means of memetic algorithms and their combination with genetic algorithms and differential evolution is the main objective of this subproject. We will pay attention their scalability for solving large scale optimization problems and their application to different MIBISCOC problems.
Vision-based assisted flexible endoscopy (Henesis, Italy)
The overall aims of the subproject are manifold, i.e. to: (i) advance the state of the art in segmentation and tracking of anatomical structures in blurred endoscopic images and sequences; (ii) advance the state of the art in vision-based endoscopic diagnosis (ex: detection of pre-cancerous tissues); (iii) apply and validate the developed techniques in realistic clinical endoscopic applications.
The subproject can be split into several phases, which are:
Multimodal behavioural assessment and prediction in rehabilitation (Henesis, Italy)
The overall aims of the subproject are manifold, i.e. to: (i) investigate and conceive Soft and Bio-inspired Computing techniques for effective and robust multimodal sensory fusion; (ii) investigate the application of bio-inspired soft computing to real-time multimodal and multi-sensor data with prediction capabilities; (iii) step forward in behavioural assessment and prediction in outdoor environment for quantitative and qualitative monitoring of a rehabilitation process, with particular emphasis on outdoor sport and post-surgical rehabilitation.
The subproject can be split into several phases, which are:
Development of novel image analysis techniques for preclinical and clinical studies, involving the problems of segmentation and non-rigid registration in real time medical imaging data (Universitätsklinikum Freiburg, Germany)
The subproject is focused on the development of novel image analysis techniques for preclinical and clinical studies, involving the problems of segmentation and non-rigid registration in real time medical imaging data. In such cases the two problems are often closely associated and can benefit from their joint treatment. The voxel classification and segmentation can be improved by considering the voxel correspondences to temporally adjacent frames and vice versa. An example is the analysis of phase contrast cardiac data for the segmentation of the myocardium and the tracking of its motion. In images where the tissue contrast also varies with time the joint treatment of segmentation and registration becomes a necessity. Its solution requires the estimation of a temporal profile for tissue contrast. An example is abdominal contrast enhanced MRI. The data requires compensation for breathing motion and segmentation for the localization of possible lesions. The contrast enhancement varies together with the time that lapsed after the injection of the contrast agent. It also varies depending on whether the tissue is healthy or pathologic.
Development of novel image analysis techniques for preclinical and clinical studies, monitoring of preclinical or clinical pharmaceutical studies with imaging (Universitätsklinikum Freiburg, Germany)
This subproject will investigate the development of novel image analysis techniques for preclinical and clinical studies, involving the monitoring of preclinical or clinical pharmaceutical studies with imaging. Such data may involve the fusion of images of the same subject acquired with different MRI contrast mechanisms and even with different imaging modalities. The monitoring is also performed longitudinally along time. The intra-subject fusion can require non-rigid registration to account for subject motion or possible geometric acquisition artefacts. Images of different time points can be analyzed independently with segmentation or with the extraction of other features. Efficient ways to incorporate user adjustments to improve the analysis of the images will also be examined. The longitudinal information will be used to derive new and improved biomarkers. The methods developed will involve extensive use of graphical models. They will also involve the use of prior geometrical or statistical modelling.