Anna Jenul holds a master’s degree in mathematics from the University of Klagenfurt and currently pursues a PhD degree in data science at the Norwegian University of Life Sciences (NMBU). In her research, she develops machine learning algorithms and statistical methods that focus on analysis of healthcare data such as treatment outcome prediction of cancer patients. Her research interests include feature selection, multiblock methods, as well as Bayesian and computational statistics.
Technical Vision Talk: “Data science for treatment outcome prediction: towards interpretable models combining healthcare data from multiple sources“
Recent advances in the field of data science offer new possibilities in the medical and healthcare sector by enhancing outcome prediction and treatment selection in the clinic. However, the need for interpretable high performance models requires improvement of existing approaches. Typical characteristics of healthcare data include high dimensionality along with low numbers of patients, leading to an increased risk of overfitting models. Adequate feature selection techniques are therefore necessary to restrict the dataset to the most informative columns, resulting in reduced complexity of machine learning models and enhanced data interpretability.
Another issue arises from the fact that information is acquired from multiple sources such as clinical data, various medical imaging modalities and gene expression data. To analyze data from various sources, we deploy extensions of well-established component based methods like PCA or PLS to unsupervised and supervised multi-block scenarios, where each block corresponds to a data source. Such multi-block methods reduce redundancies between blocks when training prediction models and simultaneously explore the underlying data structure. The combination of feature selection and multi-block methods presents a promising approach in healthcare, especially in cases where the number of patients is smaller than the number of features.
__________________
Fri. Oct 1 | 11:00 am – Technical Vision Talk: “Data science for treatment outcome prediction: towards interpretable models combining healthcare data from multiple sources“