LV 709.049 Biomedical Informatics – Discovering Knowledge in (Big*) Data
Winter Term 2016/17 – class started on October, 12, 2016, 11:15
First exam on 1th Februrary 2017; Second exam in June, Third exam in October 2017
*) and sometimes small amounts of complex data
This course covers data science aspects of biomedical informatics (= BIOinformatics + MEDICAL informatics). The focus is on knowledge discovery from complex life science data using applied machine learning and artificial intelligence.
The course is available online via the TU Graz TUbe
The class is taking place from 11:15 to 12:45 in Lecture Hall BMTEG 138, Stremayrgasse 16, EG
Tutor: Markus Plass (Holzinger-Group)
Note: Sample exam questions with solutions can be found in the Springer textbook available at the Library: Andreas Holzinger (2014). Biomedical Informatics: Discovering Knowledge in Big Data, New York: Springer. DOI: 10.1007/978-3-319-04528-3 – however, please always check the latest course material for new stuff.
This course is compulsory for bachelor students of Biomedical Engineering (5th semester), and elective for master students Biomedical Engineering, but also recommendable for Telematics (Subject area catalog: Medical Informatics, Bioinformatics, and Neuroinformatics), Informatics (Computer Science) and students from Software Development. The complete course is presented in English as it is part of the Doctoral School Informatics – so PhD students and international students are very welcome!
Definition: Biomedical Informatics (more generally called Health Informatics, HI) can be defined as an interdisciplinary field that studies and pursues the effective use of biomedical data, information and knowledge for problem solving and decision making, motivated by efforts to improve human health and well-being.
Consequently, this course is focusing on data, information and knowledge in the context of biomedicine, life sciences, health and well-being. The lectures follow a research-based teaching (RBT) style, showing the students state-of-the-art science and engineering, and discussing some underlying fundamentals and basic principles of this extremely important, challenging and future oriented field: towards the approach of personalized, molecular medicine, smart health and on how sophisticated machine learning algorithms and knowledge discovery methods can help. A grand goal of future medicine is in modelling the complexity of patients to tailor medical decisions, health practices and therapies to the indiviudal patient. This trend towards personalized medicine produces unprecedented amounts of data, see A. Holzinger, “Trends in Interactive Knowledge Discovery for Personalized Medicine: Cognitive Science meets Machine Learning“, IEEE Intelligent Informatics Bulletin, vol. 15, iss. 1, pp. 6-14, 2014.
This course will foster an integrated approach: for the successful application of machine learning algorithms in health, a comprehensive and overarching overview of the data science ecosystem and knowledge extraction and discovery pipeline is essential. This means that a multidisciplinary skill set is required, cross-domain, encompassing the following seven specialisations: 1) data science, 2) machine learning algorithms, 3) network science, 4) graphs/topology, 5) time/entropy, 6) data visualization and visual analytics, and last but not least 7) privacy, data protection, safetey and security. See the HCI-KDD approach.
Always remember: Science is to test crazy ideas – Engineering is to put these ideas into Business!
The course 2016 consists of the following 12 lectures:
- Introduction: Computer Science meets Life Sciences, challenges and problems;
- Back to the future: Fundamentals of biomedical Data, Information and Knowledge, Entropy and Kullback-Leibler Divergence;
- Structured Data: Knowledge Representation, Ontologies and Medical Classification;
- Decision, Cognition, Uncertainty, Bayesian Statistics and Probabilistic Modelling;
- Probabilistic Graphical Models Part 1: From Knowledge Representation to Graph Model Learning;
- Probabilistic Graphical Models Part 2: From Bayesian Networks to Graph Bandits;
- Dimensionality Reduction and Subspace Clustering with the Doctor-in-the-Loop;
- Biomedical Decision Making: Reasoning and Decision Support;
- Intelligent, interactive Information Visualization and Visual Analytics;
- Biomedical Information Systems and Medical Knowledge Management;
- Biomedical Data Protection: Privacy, Safety and Security;
- Summary and future challenges in biomedical informatics;
Please NOTE: Each year this course will be updated. Old lecture slides from last year can be found below, however, for the next exam only the most recent one are relevant.
| DAY and DATE | TIME | LECTURE TOPICS | SLIDES 6 per page | SLIDES with notes | SLIDES full size | VIDEO TU Graz |
| Mi, 14.10.2015 | 11:15 – 12:45 | 01 Intoduction and Overview: Computer Science meets Life Sciences, challenges and future directions |  | | |  |
| Mi, 21.10.2015 | 11:15 – 12:45 | 02 Back to the Future – Fundamentals of biomedical Data, Information and Knowledge | | | | |
| Mi, 28.10.2015 | 11:15 – 12:45 | 03 Structured Data – Coding, Classification (ICD, SNOMED, MesH, UMLS & Co) | | | | |
| Mi, 04.11.2015 | 11:15 – 12:45 | 04 Biomedical Databases: Data Acquisition, Storage, Information Retrieval and Use | | | | |
| Mi, 11.11.2015 | 11:15 – 12:45 | 05 Semi Structured, weakly structured data; Graphs, Networks and Homologies | | | | |
| Mi, 18.11.2015 | 11:15 – 12:45 | 06 Multimedia Data Mining and Knowledge Discovery | | | | |
| Mi, 25.11.2015 | 11:15 – 12:45 | 07 Knowledge, Decision, Cognition, Probability, Uncertainty, Bayesian Statistics, Probabilistic Modelling | | | | |
| Mi, 02.12.2015 | 11:15 – 12:45 | 08 Biomedical Decision Making: Reasoning and Decision Support | | | | |
| Mi, 09.12. 2015 | 11:15 – 12:45 | 09 Intelligent, Interactive Visualization and Visual Analytics | | | | |
| Mi, 16.12.2015 | 11:15 – 12:45 | 10 Biomedical Information Systems and Medical Knowledge Management | | | | |
| Mi, 13.01.2016 | 11:15 – 12:45 | 11 Biomedical Data Protection, Privacy, Safety and Security | | | | |
| Mi, 20.01.2016 | 11:15 – 12:45 | 12 Methodology for Information Systems: Usability, Quality, Benchmarking and Evaluation | | | |
The course 2015 consisted of the following 12 lectures:
- Introduction: Computer Science meets Life Sciences, challenges and future directions;
- Back to the future: Fundamentals of biomedical Data, Information and Knowledge;
- Structured Data: Coding, Classification (ICD, SNOMED, MeSH, UMLS);
- Biomedical Databases: Acquisition, Storage, Information Retrieval and Use;
- Semi structured and weakly structured data (structural homologies);
- Multimedia Data Mining and Knowledge Discovery;
- Knowledge, Decision, Cognition, Probability, Uncertainty, Bayesian Statistics, Probabilistic Modelling;
- Biomedical Decision Making: Reasoning and Decision Support;
- Interactive Information Visualization and Visual Analytics;
- Biomedical Information Systems and Medical Knowledge Management;
- Biomedical Data Protection: Privacy, Safety and Security;
- Methodology for Information Systems: Systems Design, Usability and Evaluation;
Lecture Slides from previous courses are available here:
[Lecture LV 444.152 MEDICAL INFORMATICS [2 VO WS] at the Institute of Genomics and Bioinformatics]
[Lecture LV 444.152 MEDICAL INFORMATICS [2 VO WS] at itunes TU Graz]
Exam Example is available here:
Exam Example LV 444.152 MEDICAL INFORMATICS [2 VO WS]
Springer Textbook available to this course
The course corresponds to the NEW Springer Student Textbook (available via the TU Bibilothek):
Biomedical Informatics – Discovering Knowledge in Big Data (Paperpack)
Biomedical Informatics – Discovering Knowledge in Big Data (Kindle Edition)
A more comprehensive book is the previous version (also available via the TU Bibilothek):
Biomedical Informatics: Computational Sciences meets Life Sciences (Paperpack)
Biomedical Informatics: Computational Sciences meets Life Sciences (Kindle Edition)
The life sciences, biomedicine and health are increasingly turning into a data science, where we face not only increased volumes and a diversity of highly complex, multi-dimensional and often weakly-structured and noisy data, but also the growing need for integrative machine learning approaches. Automatic Machine Learning (aML) can be of great help here, particularly when having big data sets – where algorithms can learn from it. However, sometimes we deal not with big data, but with complex data, rare events or even NP-hard problems, e.g. in subspace clustering, protein folding, or k-Anoynmization, where such aML-approaches fail or at least carries the danger of modelling artefacts. In such situations it is benefical to make use of interactive Machine Learning (iML) by putting the doctor-into-the-loop of the machine learning algorithms.