Bernd Reusch

Generation of Prime Implicants from Subfunctions and a Unifying Approach to the Covering Problem

A new method for computing the prime implicants of a Boolean function from an arbitrary sum-of-products form is given. It depends on the observation that the prime implicants of a Boolean function can be obtained from the prime implicants of its subfunctions with respect to a fixed but arbitrary variable. The problem of obtaining all irredundant sums from the list of all prime implicants and an arbitrary list of implicants representing the function is solved. The irredundant sums are in one-to-one relation to the prime implicants of a positive Boolean function associated with these lists. The known formulas of Petrick, Ghazala, Tison, Mott, and Chang are obtained as special cases and incompletely specified functions can also be handled. We give a complete and simple method for finding the positive Boolean function mentioned above. The paper is self-contained and examples are included.

Using vague knowledge for image description

Fuzzy logic in the field of image processing is becoming more and more important. Image quality is improving, but data contains different kinds of uncertainty which have to be handled using computerized analysis. These uncertainties can be found particularly in medical image data. In this paper, the handling of such uncertainties using methods such as the natural language of medical experts and iconic fuzzy sets for the description and recognition of anatomical structures in magnetic resonance tomograms (MRT) of human heads are presented.

An approach to use linguistic and model-based fuzzy expert knowledge for the analysis of MRT images

Abstract Processing of medical image data is much more difficult than industrial image processing resulting from the higher degree of variability. However the improvement of the already existing technique and the development of new imaging techniques as well as the widespread availability of recording devices (e.g. CTs and MRTs) emphasize the importance of this application field worldwide. Although the image quality increases steadily, data contains different kinds of uncertainties and variabilities that have to be handled in computerized analysis. In this paper the handling of such kinds of variabilities by using iconic fuzzy sets for the description and recognition of anatomical structures in magnetic resonance tomograms (MRT, MRI) of human heads will be presented.

Directed mutation operators – an overview

Directed mutation has shown to improve the efficiency of evolutionary algorithms significantly for a broad spectrum of optimization problems. When the first mutation operators of this kind, however, suffered from the asymmetry parameter influencing the mutation strength, in the meantime there are several new directed mutation operators available which overcome this drawback. The aim of this paper is to give an overview of all different operators in one single place. Their characteristics will be presented and their advantages and disadvantages are discussed. At the end a comparison and a summary is provided.

New model for medical image analysis based on computational intelligence

Medical image data contain several types of uncertain or vague information. Therefore, the knowledge about the data is also vague. Consequently, for the segmentation and classification of the image data the use of vague knowledge should be allowed. Finally, the optimization of such systems dealing with vague information should also be done automatically. Using fuzzy descriptions for the segmentation and analysis of medical image data has provided better results than the exclusive use of standard methods. More often, structures can be segmented and classified more adequately. Applications have been developed to segment brain structures and tumors in MRI- data. By applying neural networks, good results for the step of image improvement (pre-processing) and determination of regions of interest (ROI) in image data were obtained. Even more the classification of structures with neural networks has shown also good results. The optimization of the knowledge base was done with evolution strategies. Therefore, the optimization time was reduced to a fraction of the time needed for a manual optimization.© (2000) COPYRIGHT SPIE--The International Society for Optical Engineering. Downloading of the abstract is permitted for personal use only.

Automatische Detektion und Auswertung von Verkalkungen der Koronararterien in EBCT- und Spiral-CT-Bilddaten

Vorgestellt wird ein Bildanalysesystem fur die Auswertung von EBCT- und Spiral-CT-Bilddaten des Herzens. Das System erlaubt die Verarbeitung von unscharfem und vagem Wissen, welches sich extern in einer vom medizinischen Experten pflegbaren Wissensbasis befindet. In dieser ist der anatomische Aufbau des Herzens mit bekannten Variabilitaten in einer Sprache beschrieben, die eine einfache Erweiterung und eine leichte Einarbeitung auch fur Nicht-Informatiker ermoglicht. Die Form umgangssprachlicher Beschreibungen erlaubt die unscharfe Modellierung des Expertenwissens. Fur die Wissensreprasentation werden Frames verwendet.

Linear Realization of Finite Automata.

This paper presents a method for linear realization of automata when such exists.The basic tools employed are indexed partitions (as a way of writing certain mappings) and predecessor operations. The indexed partitions are studied in some detail and their connection to codes is explained. The method itself is simple and straightforward and contains no search steps. Only standard techniques of linear algebra, such as solving linear equations, are used. Examples are given to show the advantage of our method compared with previously known ones.

Note on Minimal Congruences on Transition Graphs

A new proof is given for the following fact: All minimal congruences on a transition graph are of one of four types and can easily be obtained by inspecting the graph.

The Influence of Lotfi Zadeh on Informatik I in Dortmund

“Around 1970 I met Lotfi for the first time — young, innocent, and as completely unknown as I was. When I reached his office, there was a note for me, ‘Sorry, forgot my birthday. Please meet me at (some Bay-Shore restaurant)’. Meeting him at his birthday party I was even more impressed by him than by the amount of shrimps I was able to swallow. He has always been that way, open to new ideas and people, helpful in the extreme, and a human in the best sense of the word. Thank you Lotfi, and thank you Fay for maintaining him so well.”

System for renal movement elimination and renal diagnosis supported by vague knowledge

For the analysis of renal function, sequences of 90 magnet resonance images of the abdominal region showing both kidneys are taken in intervals of two seconds after a contrast medium was applied. Respiration of the patients during the acquisition of the images leads to organ movements throughout the series. These displacements are corrected by using an extended cepstral technique. To minimize registration errors caused by inhomogeneous movements of organs and tissues during respiration, the cepstrum-relevant part of the images is limited to small regions of interest around both kidneys. Even organ movements of sub-pixel range can be detected. After correction, the kidneys are the same position throughout the sequence. The regions of interest marked in one image are projected to all other images. To archive diagnostic results, dynamic contrast medium evaluations for different tissues of the kidneys are computed with signal-intensity-time graphs. Using a-priori knowledge about parameters of the SIT-graph for a whole kidney and about organ shape and structure, pixels of the kidney-segment are divided into the three classes renal cortex, medulla and pelvis. As a result, precise graphs can be computed for each tissue. The evaluation of the system is in progress, time save is more than one hour per patient.