Refine
Year of publication
Document Type
- Article (19)
- Working Paper (5)
- Part of a Book (4)
- Conference Proceeding (4)
- Preprint (3)
- Report (3)
Has Fulltext
- yes (38)
Is part of the Bibliography
- no (38)
Keywords
Institute
- Informatik (38)
- Medizin (3)
- E-Finance Lab e.V. (1)
In intensive care units physicians are aware of a high lethality rate of septic shock patients. In this contribution we present typical problems and results of a retrospective, data driven analysis based on two neural network methods applied on the data of two clinical studies. Our approach includes necessary steps of data mining, i.e. building up a data base, cleaning and preprocessing the data and finally choosing an adequate analysis for the medical patient data. We chose two architectures based on supervised neural networks. The patient data is classified into two classes (survived and deceased) by a diagnosis based either on the black-box approach of a growing RBF network and otherwise on a second network which can be used to explain its diagnosis by human-understandable diagnostic rules. The advantages and drawbacks of these classification methods for an early warning system are discussed.
It is well known that artificial neural nets can be used as approximators of any continuous functions to any desired degree and therefore be used e.g. in high - speed, real-time process control. Nevertheless, for a given application and a given network architecture the non-trivial task remains to determine the necessary number of neurons and the necessary accuracy (number of bits) per weight for a satisfactory operation which are critical issues in VLSI and computer implementations of nontrivial tasks. In this paper the accuracy of the weights and the number of neurons are seen as general system parameters which determine the maximal approximation error by the absolute amount and the relative distribution of information contained in the network. We define as the error-bounded network descriptional complexity the minimal number of bits for a class of approximation networks which show a certain approximation error and achieve the conditions for this goal by the new principle of optimal information distribution. For two examples, a simple linear approximation of a non-linear, quadratic function and a non-linear approximation of the inverse kinematic transformation used in robot manipulator control, the principle of optimal information distribution gives the the optimal number of neurons and the resolutions of the variables, i.e. the minimal amount of storage for the neural net. Keywords: Kolmogorov complexity, e-Entropy, rate-distortion theory, approximation networks, information distribution, weight resolutions, Kohonen mapping, robot control.
It is well known that artificial neural nets can be used as approximators of any continous functions to any desired degree. Nevertheless, for a given application and a given network architecture the non-trivial task rests to determine the necessary number of neurons and the necessary accuracy (number of bits) per weight for a satisfactory operation. In this paper the problem is treated by an information theoretic approach. The values for the weights and thresholds in the approximator network are determined analytically. Furthermore, the accuracy of the weights and the number of neurons are seen as general system parameters which determine the the maximal output information (i.e. the approximation error) by the absolute amount and the relative distribution of information contained in the network. A new principle of optimal information distribution is proposed and the conditions for the optimal system parameters are derived. For the simple, instructive example of a linear approximation of a non-linear, quadratic function, the principle of optimal information distribution gives the the optimal system parameters, i.e. the number of neurons and the different resolutions of the variables.
One of the most interesting domains of feedforward networks is the processing of sensor signals. There do exist some networks which extract most of the information by implementing the maximum entropy principle for Gaussian sources. This is done by transforming input patterns to the base of eigenvectors of the input autocorrelation matrix with the biggest eigenvalues. The basic building block of these networks is the linear neuron, learning with the Oja learning rule. Nevertheless, some researchers in pattern recognition theory claim that for pattern recognition and classification clustering transformations are needed which reduce the intra-class entropy. This leads to stable, reliable features and is implemented for Gaussian sources by a linear transformation using the eigenvectors with the smallest eigenvalues. In another paper (Brause 1992) it is shown that the basic building block for such a transformation can be implemented by a linear neuron using an Anti-Hebb rule and restricted weights. This paper shows the analog VLSI design for such a building block, using standard modules of multiplication and addition. The most tedious problem in this VLSI-application is the design of an analog vector normalization circuitry. It can be shown that the standard approaches of weight summation will not give the convergence to the eigenvectors for a proper feature transformation. To avoid this problem, our design differs significantly from the standard approaches by computing the real Euclidean norm. Keywords: minimum entropy, principal component analysis, VLSI, neural networks, surface approximation, cluster transformation, weight normalization circuit.
The paper focuses on the division of the sensor field into subsets of sensor events and proposes the linear transformation with the smallest achievable error for reproduction: the transform coding approach using the principal component analysis (PCA). For the implementation of the PCA, this paper introduces a new symmetrical, lateral inhibited neural network model, proposes an objective function for it and deduces the corresponding learning rules. The necessary conditions for the learning rate and the inhibition parameter for balancing the crosscorrelations vs. the autocorrelations are computed. The simulation reveals that an increasing inhibition can speed up the convergence process in the beginning slightly. In the remaining paper, the application of the network in picture encoding is discussed. Here, the use of non-completely connected networks for the self-organized formation of templates in cellular neural networks is shown. It turns out that the self-organizing Kohonen map is just the non-linear, first order approximation of a general self-organizing scheme. Hereby, the classical transform picture coding is changed to a parallel, local model of linear transformation by locally changing sets of self-organized eigenvector projections with overlapping input receptive fields. This approach favors an effective, cheap implementation of sensor encoding directly on the sensor chip. Keywords: Transform coding, Principal component analysis, Lateral inhibited network, Cellular neural network, Kohonen map, Self-organized eigenvector jets.
We present a framework for the self-organized formation of high level learning by a statistical preprocessing of features. The paper focuses first on the formation of the features in the context of layers of feature processing units as a kind of resource-restricted associative multiresolution learning We clame that such an architecture must reach maturity by basic statistical proportions, optimizing the information processing capabilities of each layer. The final symbolic output is learned by pure association of features of different levels and kind of sensorial input. Finally, we also show that common error-correction learning for motor skills can be accomplished also by non-specific associative learning. Keywords: feedforward network layers, maximal information gain, restricted Hebbian learning, cellular neural nets, evolutionary associative learning
After a short introduction into traditional image transform coding, multirate systems and multiscale signal coding the paper focuses on the subject of image encoding by a neural network. Taking also noise into account a network model is proposed which not only learns the optimal localized basis functions for the transform but also learns to implement a whitening filter by multi-resolution encoding. A simulation showing the multi-resolution capabilitys concludes the contribution.
This paper describes the use of a radial basis function (RBF) neural network. It approximates the process parameters for the extrusion of a rubber profile used in tyre production. After introducing the problem, we describe the RBF net algorithm and the modeling of the industrial problem. The algorithm shows good results even using only a few training samples. It turns out that the „curse of dimensions“ plays an important role in the model. The paper concludes by a discussion of possible systematic error influences and improvements.
For the efficient management of large image databases, the automated characterization of images and the usage of that characterization for searching and ordering tasks is highly desirable. The purpose of the project SEMACODE is to combine the still unsolved problem of content-oriented characterization of images with scale-invariant object recognition and modelbased compression methods. To achieve this goal, existing techniques as well as new concepts related to pattern matching, image encoding, and image compression are examined. The resulting methods are integrated in a common framework with the aid of a content-oriented conception. For the application, an image database at the library of the university of Frankfurt/Main (StUB; about 60000 images), the required operations are developed. The search and query interfaces are defined in close cooperation with the StUB project “Digitized Colonial Picture Library”. This report describes the fundamentals and first results of the image encoding and object recognition algorithms developed within the scope of the project.
The prevention of credit card fraud is an important application for prediction techniques. One major obstacle for using neural network training techniques is the high necessary diagnostic quality: Since only one financial transaction of a thousand is invalid no prediction success less than 99.9% is acceptable. Due to these credit card transaction proportions complete new concepts had to be developed and tested on real credit card data. This paper shows how advanced data mining techniques and neural network algorithm can be combined successfully to obtain a high fraud coverage combined with a low false alarm rate.
Classically, encoding of images by only a few, important components is done by the Principal Component Analysis (PCA). Recently, a data analysis tool called Independent Component Analysis (ICA) for the separation of independent influences in signals has found strong interest in the neural network community. This approach has also been applied to images. Whereas the approach assumes continuous source channels mixed up to the same number of channels by a mixing matrix, we assume that images are composed by only a few image primitives. This means that for images we have less sources than pixels. Additionally, in order to reduce unimportant information, we aim only for the most important source patterns with the highest occurrence probabilities or biggest information called „Principal Independent Components (PIC)“. For the example of a synthetic picture composed by characters this idea gives us the most important ones. Nevertheless, for natural images where no a-priori probabilities can be computed this does not lead to an acceptable reproduction error. Combining the traditional principal component criteria of PCA with the independence property of ICA we obtain a better encoding. It turns out that this definition of PIC implements the classical demand of Shannon’s rate distortion theory.
The efficient management of large multimedia databases requires the development of new techniques to process, characterize, and search for multimedia objects. Especially in the case of image data, the rapidly growing amount of documents prohibits a manual description of the images’ content. Instead, the automated characterization is highly desirable to support annotation and retrieval of digital images. However, this is a very complex and still unsolved task. To contribute to a solution of this problem, we have developed a mechanism for recognizing objects in images based on the query by example paradigm. Therefore, the most salient image features of an example image representing the searched object are extracted to obtain a scale-invariant object model. The use of this model provides an efficient and robust strategy for recognizing objects in images independently of their size. Further applications of the mechanism are classical recognition tasks such as scene decomposition or object tracking in video sequences.
At present, there are no quantitative, objective methods for diagnosing the Parkinson disease. Existing methods of quantitative analysis by myograms suffer by inaccuracy and patient strain; electronic tablet analysis is limited to the visible drawing, not including the writing forces and hand movements. In our paper we show how handwriting analysis can be obtained by a new electronic pen and new features of the recorded signals. This gives good results for diagnostics. Keywords: Parkinson diagnosis, electronic pen, automatic handwriting analysis
Erkennung kritischer Zustände von Patienten mit der Diagnose "Septischer Schock" mit einem RBF-Netz
(2000)
Es wurde gezeigt, dass der Arzt mit dem wachsenden RBF-Netz durch die Ausgabe von verlässlichen Warnungen unterstützt werden kann. Wie in der Clusteranalyse erläutert, leiden die Ergebnisse jedoch unter den wenigen Patienten und unter der ungenauen zeitlichen Erfassung der Daten. Da jeder Patient sehr individuelle Zustände annimmt, ist ein größeres Patientenkollektiv notwendig, um eine umfassende Wissensbasis zu lernen. Eine medizinische Nachbearbeitung der Wissensbasis durch die Analyse der Fälle ließe eine weitere Verbesserung des Ergebnisses erwarten. Somit könnten unbekannte Zusammenhänge durch das Lernen aus Beispielen und medizinisches Fachwissen kombiniert werden. Abstraktere Merkmale, die weniger abhängig von individuellen Zuständen sind, könnten eine Klassifikation noch weiter verbessern. Ein Ansatzpunkt ist z.B. die Abweichung der Messwerte vom gleitenden Mittelwert. Dieses Maß ist unempfindlicher gegenüber den individuellen Arbeitspunkten der Patienten und bildet auch die Basis von relativen Abhängigkeiten zwischen zwei Variablen, die in einem weiteren Schritt ebenfalls als Merkmal herangezogen wurden. Obwohl die Verwendung der relativen Abhängigkeiten zwischen zwei Variablen als Merkmal nicht deutlichere oder häufigere Warnungen hervorbringen konnte, weist doch die Clusteranalyse auf eine bessere Verteilung der Patienten hin. Einige Cluster sind besser für die Vorhersage geeignet, als dieses bei einer Clusterung auf Basis der Zustände erreicht werden kann. Unterstützt wird dieses Ergebnis auch durch den größeren Unterschied der Sicherheiten von falschen und richtigen Klassifikationen. Neben den bisher untersuchten Merkmalen scheinen auch die Variablen interessant zu sein, bei denen festgestellt wurde, dass sie sich trotz Medikamentengabe und adäquater Behandlung schwer stabilisieren lassen. Durch den behandelnden Arzt werden diese Werte üblicherweise in einem gewissen Bereich gehalten. Falls sich das Paar Medikament/physiologischer Parameter nicht mehr in einem sinnvollen Verhältnis befindet, kann dieses ein wichtiger Indikator sein. Nach dem Aufbau der grundlegenden Funktionalität der hier untersuchten Methoden ist die Suche nach geeigneten Merkmalen als Eingabe für ein neuronales Netz ein wesentlicher Bestandteil folgender Arbeiten. Abgesehen von dem generell anspruchsvollen Vorhaben aus Klinikdaten deutliche Hinweise für die Mortalität septischer-Schock-Patienten zu erhalten, liegen die wesentlichen Probleme in dem Umfang und der Messhäufigkeit der Frankfurter Vorstudie begründet, so dass eine Anwendung von Klassifikationsverfahren auf das umfassendere Patientenkollektiv der MEDAN Multicenter-Studie klarere Ergebnisse erwarten lässt. Eine weitere, für medizinische Anwendungen interessante, Analysemöglichkeit ist die Regelgenerierung, die zur Zeit in einem anderen Teilprojekt in der MEDAN-Arbeitsgruppe bearbeitet wird. Hier können im Fall metrischer Daten zusätzliche Hinweise für die Leistung eines reinen Klassifikationsverfahrens gewonnen werden mit dem Vorteil einer expliziten Regelausgabe. Zum anderen werden in diesem Teilprojekt auch Verfahren zur Regelgenerierung eingesetzt, die ordinale und nominale Variablen wie Diagnosen, Operationen, Therapien und Medikamentenangaben (binär, ohne genaue Dosis) auswerten können. Diese werden in den Multicenter-Daten vorhanden sein. Durch Kopplung der Regelgeneratoren für metrische Daten auf der einen Seite und für diskrete Variablen auf der anderen Seite, besteht durchaus die Hoffnung bessere Ergebnisse zu erzielen. Da der Regelgenerator für metrische Daten auf dem RBF-DDA (Abk. für: Dynamic Decay Adjustment)-Netz [BERTHOLD und DIAMOND, 1995] beruht, bietet es sich innerhalb des MEDAN-Projekts an, einen (bislang nicht durchgeführten) Vergleich mit dem hier verwendeten Netztyp durchzuführen. Der Vergleich ist allerdings nur von prinzipiellem Interesse und kann auf den hier betrachteten Daten kein grundsätzlich besseres Ergebnis liefern als die bislang durchgeführten Analysen; er kann aber zu einer umfangreichen Bewertung der Ergebnisse beitragen.
Since the description of sepsis by Schottmüller in 1914, the amount on knowledge available on sepsis and its underlying pathophysiology has substantially increased. Epidemiologic examinations of abdominal septic shock patients show the potential for high risk posed by and the extensive therapy situation in the intensive care unit (ICU) (5). Unfortunately, until now it has not been possible to significantly reduce the mortality rate of septic shock, which is as high as 50-60% worldwide, although PROWESS' results (1) are encouraging. This paper summarizes the main results of the MEDAN project and their medical impacts. Several aspects are already published, see the references. The heterogeneity of patient groups and the variations in therapy strategies is seen as one of the main problems for sepsis trials. In the MEDAN multi-center study of 71 intensive care units in Germany, a group of 382 patients made up exclusively of abdominal septic shock patients who met the consensus criteria for septic shock (3) was analysed. For use within scores or stand-alone experiments variables are often studied as isolated variables, not as a multidimensional whole, e.g. a recent study takes a look at the role thrombocytes play (15). To avoid this limitation, our study compares several established scores (SOFA, APACHE II, SAPS II, MODS) by a multi-dimensional neuronal network analysis. For outcome prediction the data of 382 patients was analysed by using most of the commonly documented vital parameters and doses of medicine (metric variables). Data was collected in German hospitals from 1998 to 2001. The 382 handwritten patient records were transferred to an electronic database giving the amount of 2.5 million data entries. The metric data contained in the database is composed of daily measurements and doses of medicine. We used range and plausibility checks to allow no faulty data in the electronic database. 187 of the 382 patients are deceased (49 %).
Im Zeitraum 1. 11. 1993 bis 30. 3. 1997 wurden 1149 allgemeinchirurgische Intensivpatienten prospektiv erfaßt, von denen 114 die Kriterien des septischen Schocks erfüllten. Die Letalität der Patienten mit einem septischen Schock betrug 47,3%. Nach Training eines neuronalen Netzes mit 91 (von insgesamt n = 114) Patienten ergab die Testung bei den verbleibenden 23 Patienten bei der Berücksichtigung von Parameterveränderungen vom 1. auf den 2. Tag des septischen Schocks folgendes Ergebnis: Alle 10 verstorbenen Patienten wurden korrekt als nicht überlebend vorhergesagt, von den 13 Überlebenden wurden 12 korrekt als überlebend vorhergesagt (Sensitivität 100%; Spezifität 92,3%).