We propose a new deformable model Deformable Associate Net (DAN). It is represented by a set of nodes which are associated by deformation constrains such as topology association, inter-part association, intra-part association, and geometry to atlas association. Each node in the model is given a priority, and hence DAN is a hierarchical model in which each layer is decided by nodes with same priority. Directional edges and dynamic generated local atlases are used in energy function to incorporate knowledge about tissue and image acquisition. A fast digital topology based method is designed to check whether topology of the model is changed under deformation. The deformation procedure hierarchically combines global and local deformations. Layers with high priority deform first. Once a higher layer is deformed to its target position in an image, the nodes in this layer are fixed, and then used as reference to help lower layers deform to their initial positions. At a particular layer, the model is first deformed by using global affine transformation to fit the image roughly, and then is warped by using a local deformation to fit the image better. The proposed method has been used to segment chest CT images for thoracic surgical planning, and it is also promising for other medical applications, such as model based image registration, and model-based 3D modeling.
Our previous work to segment the complete ventricular system from T1-weighted and SPGR MR images is extended here to deal with T2-weighted MR images. For each ventricle, a region of interest is determined first and its local statistics is calculated to find the intensity ranges of cerebrospinal fluid, grey matter and white matter. Then, region growing is performed in each ventricle based on the calculated statistics. During region growing, anti-leakage conditions are checked to prevent growing into extraventricular spaces. With the incorporation of domain knowledge, radiological properties and geometrical constraints, the algorithm provides a means for the extraction of the ventricular system from T2-weighted MR images. Initial experimental results are presented with the extracted third and fourth ventricles.
We suggest a method to reconstruct intermediate slices of 3D subcortical structures given by slices (2D cross-sections) using shape-based interpolation for sparse slices. This algorithm overcomes some limitations of previous interpolation methods of reconstruction. The method can find the intermediate slice cross-section of 3D structure with discontinuities and can work for non-overlapping contours on different cross-sections. The method used can be considered as "indirect interpolation", because the auxiliary information about structure shape, based on minimum distance to the contour from each point of the image, has been used. Variations of the parameters (bias, connectivity, tension) can help adjust the contour produced to the expected shape. We apply this approach to interpolate subcortical structures in the Talairach-Tournoux brain atlas.
The human cerebral ventricular system is a complex structure that is essential for the well being and changes in which reflect disease. It is clinically imperative that the ventricular system be studied in details. For this reason computer assisted algorithms are essential to be developed. We have developed a novel (patent pending) and robust anatomical knowledge-driven algorithm for automatic extraction of the cerebral ventricular system from MRI. The algorithm is not only unique in its image processing aspect but also incorporates knowledge of neuroanatomy, radiological properties, and variability of the ventricular system. The ventricular system is divided into six 3D regions based on the anatomy and its variability. Within each ventricular region a 2D region of interest (ROI) is defined and is then further subdivided into sub-regions. Various strict conditions that detect and prevent leakage into the extra-ventricular space are specified for each sub-region based on anatomical knowledge. Each ROI is processed to calculate its local statistics, local intensity ranges of cerebrospinal fluid and grey and white matters, set a seed point within the ROI, grow region directionally in 3D, check anti-leakage conditions and correct growing if leakage occurs and connects all unconnected regions grown by relaxing growing conditions. The algorithm was tested qualitatively and quantitatively on normal and pathological MRI cases and worked well. In this paper we discuss in more detail inclusion of anatomical knowledge in the algorithm and usefulness of our approach from clinical perspective.
Cerebrospinal fluid filled ventricular system is an essential part of brain. The volume, shape and size of this ventricular system remain more or less constant and various pathologies directly or indirectly affect them. Morphometric analysis of cerebral ventricular system is important for evaluating changes due to growth, aging, intrinsic and extrinsic pathologies. Previous quantification efforts using ex vivo techniques suffered considerable error due to deformation of slices during sectioning, and numerous other factors. In vivo studies using air or contrast media also introduce volumetric changes in the ventricles thus giving erroneous quantitative information. Imaging of ventricular anatomy avoids these problems and allows repetitive studies following progression of ventricular system changes due to disease or natural processes. We have developed a methodology for automated extraction of ventricular system from MR neuroimages. Once extracted, landmarks are located on the surface of ventricular system automatically. These landmarks are then used for calculation of the ventricular shape, volume and size. A total of 20 brain ventricular systems were analyzed. The morphometric dimensions of the ventricles are presented in this paper. This study forms an initial basis for more advanced work on ventricular segmentation and morphometry.
An algorithm to automatically detect brain tumors in MR images is presented. The key concern is speed in order to process efficiently large brain image databases and provide quick outcomes in clinical setting. The method is based on study of asymmetry of the brain. Tumors cause asymmetry of the brain, so we detect brain tumors in 3D MR images using symmetry analysis of image grey levels with respect to the midsagittal plane (MSP). The MSP, separating the brain into two hemispheres, is extracted using our previously developed algorithm. By removing the background pixels, the normalized grey level histograms are calculated for both hemispheres. The similarity between these two histograms manifests the symmetry of the brain, and it is quantified by using four symmetry measures: correlation coefficient, root mean square error, integral of absolute difference (IAD), and integral of normalized absolute difference (INAD). A quantitative analysis of brain normality based on 42 patients with tumors and 55 normals is presented. The sensitivity and specificity of IAD and INAD were 83.3% and 89.1%, and 85.7% and 83.6%, respectively. The running time for each symmetry measure for a 3D 8bit MR data was between 0.1 - 0.3 seconds on a 2.4GHz CPU PC.
An efficient database is an essential component of organizing diverse information on image metadata and patient information for research in medical imaging. This paper describes the design, development and deployment of a large database system serving as a brain image repository that can be used across different platforms in various medical researches. It forms the infrastructure that links hospitals and institutions together and shares data among them. The database contains patient-, pathology-, image-, research- and management-specific data. The functionalities of the database system include image uploading, storage, indexing, downloading and sharing as well as database querying and management with security and data anonymization concerns well taken care of. The structure of database is multi-tier client-server architecture with Relational Database Management System, Security Layer, Application Layer and User Interface. Image source adapter has been developed to handle most of the popular image formats. The database has a user interface based on web browsers and is easy to handle. We have used Java programming language for its platform independency and vast function libraries. The brain image database can sort data according to clinically relevant information. This can be effectively used in research from the clinicians’ points of view. The database is suitable for validation of algorithms on large population of cases. Medical images for processing could be identified and organized based on information in image metadata. Clinical research in various pathologies can thus be performed with greater efficiency and large image repositories can be managed more effectively. The prototype of the system has been installed in a few hospitals and is working to the satisfaction of the clinicians.