Entity

Time filter

Source Type


Perner P.,Institute of Computer Vision and Applied Computer science
Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics) | Year: 2011

Data mining methods are widely used across many disciplines to identify patterns, rules or associations among huge volumes of data. While in the past mostly black box methods such as neural nets and support vector machines have been heavily used in technical domains, methods that have explanation capability are preferred in medical domains. Nowadays, data mining methods with explanation capability are also used for technical domains after more work on advantages and disadvantages of the methods has been done. Decision tree induction such as C4.5 is the most preferred method since it works well on average regardless of the data set being used. This method can easily learn a decision tree without heavy user interaction while in neural nets a lot of time is spent on training the net. Cross-validation methods can be applied to decision tree induction methods; these methods ensure that the calculated error rate comes close to the true error rate. The error rate and the particular goodness measures described in this paper are quantitative measures that provide help in understanding the quality of the model. The data collection problem with its noise problem has to be considered. Specialized accuracy measures and proper visualization methods help to understand this problem. Since decision tree induction is a supervised method, the associated data labels constitute another problem. Re-labeling should be considered after the model has been learnt. This paper also discusses how to fit the learnt model to the expert́s knowledge. The problem of comparing two decision trees in accordance with its explanation power is discussed. Finally, we summarize our methodology on interpretation of decision trees. © 2011 Springer-Verlag. Source


Perner P.,Institute of Computer Vision and Applied Computer science
Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics) | Year: 2014

Unsupervised classification is the choice when knowledge about the class numbers and the class properties is missing. However, using clustering might not lead to the correct class and needs interacting with the domain experts to figure out the classes that make sense for the respective domain. We propose to use a prototype-based learning and classification method in order to figure out the right number of classes and the class description. An expert might start with picking out a prototypical image or object for the class he is expecting. Later on, he might pick out some more prototypes that might represent the variance of the class. By doing so might be incrementally learnt the class border and the knowledge about the class. It does not need the expert so heavy interaction with the system. Such a method is especially useful when the domain has very noisy objects and images. We present in the paper the method for prototype-based classification, the methodology, and describe the success of the method on a biological application - the detection of different dynamic signatures of mitochondrial movement. © 2014 Springer International Publishing Switzerland. Source


Perner P.,Institute of Computer Vision and Applied Computer science
CEUR Workshop Proceedings | Year: 2014

An expert is able to tell the system developer in many image-related tasks what a prototypical image should look like. Usually he will choose several prototypes for one class, but he cannot provide a good and large enough sample set for the class to train a classifier. Therefore, we mapped his technical procedure into a technical system based on proper theoretical methods that assist him in acquiring the knowledge about his application and furthermore in developing a classifier for his task. This system helps him to learn about the clusters and the borderlines of the clusters even when the data are very noisy as is the case for microscopic cell images in drug discovery, where it is unclear if the drug will produce the expected result on the cell parts. We describe in this paper the necessary functions that a prototype-based classifier should have. We also use the expert's estimated similarity as a new knowledge piece and based on that we optimize the similarity. The test of the system was carried out on a new application on microscopic cell image analysis - the study of the internal mitochondrial movement of cells. The aim was to discover the different dynamic signatures of mitochondrial movement. Three results of this movement were expected: tubular, round, and dead cells. Based on our results we can show the success of the developed method. Copyright © 2014 by the paper's authors. Source


Perner P.,Institute of Computer Vision and Applied Computer science
Procedia Computer Science | Year: 2015

To study human image cognition is more than ever an important topic since the number of vision-based materials has been increased over the years. Texture seems to be a powerful tool to describe the appearances of objects. Therefore, very flexible and powerful texture descriptors are of importance that allow to recognize the texture and to understand what makes up the texture. The most used texture descriptor is the well-known texture descriptor based on the co-occurrence matrix. We propose a texture descriptor based on random sets. This descriptor gives us more freedom in describing different textures. In this paper, we compare the two texture descriptors based on a medical data set. We review the theory of the two texture descriptors and describe the procedure for the comparison of the two methods. Polyp images are used that are derived from colon examination. Decision tree induction is used to learn a classifier model. Cross-validation is used to calculate the error rate. The comparison of the two texture descriptors is based on the error rate, the properties of the two best classification models, the runtime for the feature calculation, the selected features, and the semantic meaning of the texture descriptors. The medical data set was chosen since texture seems to play an important role in describing medical objects. © 2015 The Authors. Published by Elsevier B.V. Source


Perner P.,Institute of Computer Vision and Applied Computer science
Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics) | Year: 2015

Live-cell assays are used to study the dynamic functional cellular processes in High-Content Screening (HCA) of drug discovery processes or in computational biology experiments. The large amount of image data created during the screening requires automatic image-analysis procedures that can describe these dynamic processes. One class of tasks in this application is the tracking of cells. We describe in this paper a fast and robust cell tracking algorithm applied to High-Content Screening in drug discovery or computational biology experiments. We developed a similarity-based tracking algorithm that can track the cells without an initialization phase of the parameters of the tracker. The similarity-based detection algorithm is robust enough to find similar cells although small changes in the cell morphology have been occurred. The cell tracking algorithm can track normal cells as well as mitotic cells by classifying the cells based on our previously developed texture classifier. Results for the cell path are given on a test series from a real drug discovery process. We present the path of the cell and the low-level features that describe the path of the cell. This information can be used for further image mining of high-level descriptions of the kinetics of the cells. © Springer International Publishing Switzerland 2015. Source

Discover hidden collaborations