Wido Menhardt

Computer vision on magnetic resonance images

We present an approach for the automated interpretation of transaxial cranial magnetic resonance images. After a brief outline of our notation and basic assumptions, the overall design consisting of a neurological inference engine, a set of image processing operators and a configurating component for these operators is presented.

An environment for three-dimensional shaded perspective display of cell components

Abstract An integrated environment for converting electron microscopic profile images into shaded perspective displays is described. Its application, reported here, is to the display of various cell organelles; it can also be used for other image acquisition modalities. What separates this application from other, primarily medical ones, is its complexity, since dictyosomes have been employed for the development of the display program. The surface profile structures of these organelles are among the most complicated being processed for scientific applications. Therefore, a high degree of interactivity had to be integrated into the three-dimensional image processing described.

Image analysis using iconic fuzzy sets

A formalism is presented for the iconic representation of knowledge in a semantic net. Nodes are represented as fuzzy sets, relations as functions over fuzzy sets. Moreover, the author demonstrates a consistent framework for the transition from the symbolic to the iconic level. He also defines a procedure for the explication, fuzzyfication and operationalization of natural language description predicates. This concept has successfully been used for a model-driven image analysis approach in the domain of MRI (magnetic resonance imaging) brain slices. The concept also has the potential for a data-driven approach. Measurements of features of fuzzy image structures can be represented in the semantic net as fuzzy assertions. Possible faults (pathologies) can be incorporated by using information form other sources (neurological studies) as (fuzzy) assertions.<<ETX>>

Iconic Fuzzy Sets for MR Image Segmentation

In this paper we present an approach to the top-down analysis of magnetic resonance images using the concept of iconic fuzzy sets. A symbolic model is used to describe the properties and relationships of scene objects. Using iconic fuzzy sets as pixel level representations of image objects, and fuzzy operationalizations of relations of these objects, a new framework for image analysis can be defined. We will demonstrate the feasibility of the approach with a detailed example of a supraorbital transaxial MRI brain slice.

Three-dimensional reconstruction of a plant dictyosome from series of ultrathin sections using computer image processing

A single dictyosome from an actively secreting ovary gland cell of Aptenia cordifolia has been reconstructed in 3‐D from a series of twenty‐nine electron micrographs by computer image processing. The reconstruction is presented under different viewing angles in the form of shaded perspective displays. From these displays the entire dictyosome, surrounded by numerous vesicles, appears to be more a spherical than a flat body. The plate‐like region of the dictyósome is demonstrated when only a portion of the electron micrographs is used for the image processing, leading to ‘cut‐off’ displays. Since some upper planes were removed, such ‘cut‐off’ displays revealed both tubular connections between cisternae of the dictyosome and the neighbouring endoplasmic reticulum as well as tubular continuities between adjacent Golgi cisternae within the same stack. Possible consequences of both types of interconnections on transport and processing of proteins and glycoproteins are discussed.

Wissensbasierte Konfigurierung von Interpretationsoperatoren an Hand eines hierarchischen Szenenmodells

Ein moglicher Ansatz zur Interpretation von Bildern besteht aus der schrittweisen Extraktion von Substrukturen aus bereits interpretierten Strukturen. In [1] und [2] wird ein System zur Interpretation trans-axialer kranialer MR-Bilder beschrieben, dessen Konzept gerade auf dieser Vorgehensweise beruht.

Ein Ansatz für die Interpretation von MR - Bildern

MR- (magnetic resonance-) Bilder sind Funktionen verschiedener Maschinenparameter und ortsabhangiger Gewebeparameter. Zu den Gebeparametern zahlen unter anderen die Spindichte ρ, die Spin-Gitter-Relaxationszeit T1 und die Spin-Spin-Relaxationszeit T2. Als Maschinenparameter werden Grosen wie die Repetitionszeit (TR) oder die Echozeit (TE) bezeichnet, die vom Operator spezifiziert werden konnen. So gilt fur das Signal einer Spin-Echo-Pulssequenz [1] an jedem Bildpunkt naherungsweise

Wissensbasierte Interpretation kranialer MR-Bilder

MR-Bilder sind Funktionen verschiedener Maschinenparameter und ortsabhangiger Gewebeparameter: Zu den Gewebeparametern zahlen unter anderen die Wasserstoffprotonendichte (Spindichte) rho, die Spin-Gitter-Relaxationszeit T1 und die Spin-Spin-Relaxationszeit T2; als Maschinenparameter werden Grosen wie die Repetitionszeit (TR) oder die Echozeit (TE) bezeichnet. So gilt fur das Signal einer Spin-Echo-Pulssequenz an jedem Bildpunkt naherungsweise

3-D Shaded Perspective Display of NMR-Tomograms — The Software Environment

Three dimensional perspective display consists of mathematically assigning a brightness to each point on the surface of a volume object and transmitting this brightness distribution to a video-display unit. In order to display NMR images the data must first be brought into the appropriate form. This process entails user interaction with the data. This paper concentrates on the interaction.