Medical mining is a broad research area, where mining methods are applied to solve problems of diagnostics and treatment, as well as for the understanding of disease progression. Medical mining encompasses learning on hospital records (for decision support in diagnosis and treatment), learning on healthcare-associated data, and learning on epidemiological data.
The KMD lab works mainly on epidemiological mining.
- We cooperate with the Institute of Community Medicine, University Medicine Greifswald, on the identification of risk factors and predictive factors for hepatic steatosis. In this cooperation, we study longitudinal data from the cohorts SHIP and SHIP-TREND (Study of Health in Pomerania). We develop methods for learning on high-dimensional, timestamped, multi-relational data. We address challenges of object dissimilarity, data skew and of missing information (due to changes in the recording protocol).
- Subpopulation Discovery and Validation in Epidemiological Data. EuroVis Workshop on Visual Analytics, 2017.
- Combining Subgroup Discovery and Clustering to Identify Diverse Subpopulations in Cohort Study Data. Proc. of the 30th IEEE Int. Symposium on Computer-Based Medical Systems (CBMS17), Thessaloniki, Greece, 2017.
- ICE: Interactive Classification Rule Exploration on Epidemiological Data. Proc. of the 30th IEEE Int. Symposium on Computer-Based Medical Systems (CBMS17), Thessaloniki, Greece, 2017.
- Can we classify the participants of a longitudinal epidemiological study from their previous evolution?. Proc. of the 28th IEEE Int. Symposium on Computer-Based Medical Systems (CBMS15), 121-126, IEEE, São Carlos and Ribeirão Preto, Brazil, June 2015. URL
- Mining longitudinal epidemiological data to understand a reversible disorder. Proc. of the 13th Int. Symposium on Intelligent Data Analysis (IDA'14), Springer, Leuven, Belgium, 2014.
- Using Participant Similarity for the Classification of Epidemiological Data on Hepatic Steatosis. Proc. of the 27th IEEE Int. Symposium on Computer-Based Medical Systems (CBMS14), IEEE, Mount Sinai, NY, 2014. URL
- Learning and inspecting classification rules from longitudinal epidemiological data to identify predictive features on hepatic steatosis. Expert Systems with Applications, (41)11:5405-5415, Elsevier BV, September 2014. URL
- Subpopulation Discovery in Epidemiological Data with Subspace Clustering. Foundations of Computing and Decision Sciences (FCDS), (39)4:271-300, 2014. URL
- Interactive Medical Miner: Interactively Exploring Subpopulations in Epidemiological Datasets. In Calders et al. (Eds.), European Conference on Machine Learning and Principles and Practice of Knowledge Discovery in Databases (ECML PKDD 2014) - DEMO TRACK, (8726):460-463, Springer Berlin Heidelberg, 2014. URL
- We cooperate with the Center of Biomedical Technology (CTB) and the Univ. Polytecnica de Madrid on "Data Mining and Stream Mining for Epidemiological Studies on the Human Brain" (StreaMED). We develop methods for modeling disease progression and understanding the impact of treatments.
- Predicting the post-treatment recovery of patients suffering from traumatic brain injury (TBI). Brain Informatics, 1-12, Springer Berlin Heidelberg, 2015. URL
- Are Some Brain Injury Patients Improving More Than Others?. The 2014 International Conference on Brain Informatics and Health (BIH \'14), Warsaw, Poland., 2014.
- We cooperate with the Visualization Lab of the Faculty of Computer Science on the classification of tumor lesions, using DCE-MR images. We develop methods for building regions/clusters inside each lesion, characterizing them on malignancy and using them to assess the malignancy of the tumor as a whole.
- Can we distinguish between benign and malignant breast tumors in DCE-MRI by studying a tumor's most suspect region only?. Computer-Based Medical Systems (CBMS), 2013 IEEE 26th International Symposium on, 77-82, June 2013. URL
- Classification of Benign and Malignant DCE-MRI Breast Tumors by Analyzing the Most Suspect Region. In Hans-Peter Meinzer, Thomas Martin Deserno, Heinz Handels, and Thomas Tolxdorff (Eds.), Bildverarbeitung für die Medizin 2013, 45-50, Springer Berlin Heidelberg, 2013. URL
- We cooperate with the Univ. of Ulm and the University Medicine Regensburg on methods to analyze the evolution of Tinnitus patients. We derive evolutionary patterns with respect to this chronic, still uncurable disease. Our goal is to support patients in their daily life by providing (self-)aiding measures.
Selected aspects of the KMD research on Medical Mining are partially funded through IMPRINT (model learning on multi-relational objects) and through the Innofonds of the Otto-von-Guericke University Magdeburg (preparatory work for involvement in the main phase of the Human Brain Project).