Cancer is an umbrella term for a group of diseases manifesting in more than 100 different types and subtypes, involving abnormal cell growth with the potential to invade or spread to other tissues. Most organs and tissues can be affected, while the underlying instrumental mutations and genome rearrangements are incredibly complex and often hard to detect as well as predict. Cancer is the second leading cause of death globally and it is estimated to be responsible for 9.6 million deaths (ca. 15%) in 2018 alone. Over 80% of malignant tumors are solid and the two pediatric cancers Neuroblastoma (NB, the most frequent solid cancer of early childhood) and Diffuse Intrinsic Pontine Glioma (DIPG, the leading cause of brain tumor-related death in children) are relatively representative of the cancer disease itself and have high societal impact.
As part of this research, we plan to investigate the use of novel imaging biomarkers and in-silico tumor growth simulation results to develop predictive systems for doctors and medical personal to assist diagnosis, prognosis, therapy choices, and treatment follow up. We plan to design advanced visualization methods allowing for the exploration and analysis of the multi-modal clinical data, supporting patient comparisons and common pattern identification as well as translating this knowledge into predictors for the most relevant clinical outcomes.
- What interactive visual data exploration interfaces facilitate an expert's ability to overview, analyze, and compare patient data?
- What information and capabilities are doctors requiring?
- How can individual patient biomarkers and disease courses be compared visually?
- Which novel visual encodings and interaction techniques can be used to visualize relevant clinical data?
- How can an expert's domain knowledge be leveraged and integrated into the analysis process?
- How can clinical decision making be supported?
- How can we assess the reliability of predictive data?
It is not required to complete all tasks for a single project/thesis. One should choose and restrict to a sensible (and manageable) subset. Feel free to discuss your preferences with us!
- Assessing which information and capabilities are required by doctors.
- Assessing which existing methods and visualizations exists in this field.
- Design of an interactive, visual data exploration interface that facilities experts to get an overview of the data.
- Ability to compare individual patients visually to facilitate visual reasoning with respect to patient-to-patient comparisons.
- Visually cluster patients that behave similar or dissimilar.
- Analyze how (hierarchical) clustering techniques and correlation analysis might benefit the analysis.
- Design appropriate visual encodings for clinical data and the corresponding attributes and get feedback from experts.
- Design appropriate interaction techniques to support the encodings and the analysis.
- Be able to easily identify patterns, such as positive or negative correlations between biomarkers or trends in the data, that can be indicators for a specific disease.
- Take historical data (course of the disease with the evolution of biomarkers and multi-modal data) into account and enrich current cases with historical results.
- Find similar patients using similarity search and user-defined patterns, based on bio-markers (generic disease profiles) and be able to put forward and test hypothesis
- Design measures to visualize and assess the reliability of the predictive data
- Work on cutting edge research that has important real-life implications.
- Apply computer science to the medical domain as part of a larger European research collaboration
- Ability to align topics from seminar, project, and thesis and the ability to continue work in this field with more specialized topics for a later thesis.
- Interest in the medical domain
- Highly motivated and be prepared to think outside the box
- Excellent programming skills in Python / D3 / Visualization or comparable
Due to the broad scope of the overall research, the individual work will be restricted to sub-aspects by preferences and compatibility. It is not required to complete all tasks listed above, but one should restrict the work to a sensible (and manageable) subset for the planned project duration. Starting for a bachelor's (e.g., Investigating visual encoding and overview exploration) and continuing later for a master's (e.g., Reliability assessment of trends and pattern detection in biomarker indications) is possible and encouraged as the research in this area will run for at least the next three years. Feel free to discuss your preferences with us!
- Scope: Bachelor / Master
- Project / Thesis Duration (Bachelor): 3 months + 3 months
- Project / Thesis Duration (Master): 6 months + 6 months
- Start: Planned in February 2020
- W. Aigner, P. Federico, T. Gschwandtner, S. Miksch, A. Rind (2012). Challenges of Time-oriented Data in Visual Analytics for Healthcare. Proceedings of the IEEE VisWeek Workshop on Visual Analytics in Healthcare, Seattle
- Levy-Fix, G., Kuperman, G. J., & Elhadad, N. (2019). Machine Learning and Visualization in Clinical Decision Support: Current State and Future Directions. arXiv preprint arXiv:1906.02664.
- Ltifi, H., & Ayed, M. B. (2016). Visual Intelligent Decision Support Systems in the Medical Field: Design and Evaluation. In Lecture Notes in Computer Science (pp. 243–258). DOI: 10.1007/978-3-319-50478-0_12
- Hanahan, D., & Weinberg, R. A. (2000). The Hallmarks of Cancer. Cell, 100(1), 57–70. DOI: 10.1016/s0092-8674(00)81683-9
- Hanahan, D. (2014). Rethinking the war on cancer. The Lancet, 383(9916), 558–563. DOI: 10.1016/s0140-6736(13)62226-6
- L. Gortner et al. Duale Reihe Pädiatrie, 5ed ed, Georg Thieme, 2018, ISBN 978-3-13-241153-1, p. 513-516
- D. Henne-Bruns et al. Duale Reihe Chirurgie, 5ed, Georg Thieme, 2013, ISBN 978-3-13-125294-4, p. 981-983