About the Course
This course is part of the "Open Educational Resources in Computational Biomedicine"* project proposed by NordBioMed.org and funded by Erasmus+.
NordBioMed is a collaborative network in the field of Biomedicine between the Universities of Turku, Eastern Finland, Bergen, Odense and Karolinska Institutet. The network was originally formed in 2013 to strengthen the individual biomedical teaching programs within the component universities and make them internationally more competitive by providing complementary activities from the partner universities. The network supports both student and teacher mobility, organises intensive courses and develops virtual online teaching and an information platform, NordBioMed Moodle (previously), and Open edX (upcoming).
Biomedicine covers those areas of human biology, chemistry and medicine that seek to explain the factors behind health and disease at the molecular and cellular level. This information is applied in the development of better diagnostics and treatments.
Links that redirects to the study programme pages of each NordBioMed partner universities can be found here.
You can read more about OER in this Foundations for OER Strategy Development document.
*) "Open Educational Resources in Biomedicine" - a project proposal to the Erasmus+ Strategic Partnerships call.
- Biomedical Ethics- O1 (to be based at Karolinska Institutet - KIx)
- Translational digital pathology - O2
- Introduction to Computational Biomedicine and Machine Learning - O4
- Summer School in Computational Biomedicine
- Learning Analystics - Open edX (Bergen, KI)
Introduction to Computational Biomedicine and Machine Learning
The development of computer algorithms that improve with experience by learning labelled samples, holds promise to enable computers to assist humans in the analysis of large, complex datasets such as bio-images and omics data. More recently, deep neural networks based machine learning models such as convolutional neural networks (CNN), have started an ongoing revolution in both (bio)medical imaging and integrated analysis of omics data. Deep learning was named one of “10 breakthrough technologies” by MIT Technology Review in 2013 and a “Method to watch” by Nature in 2016, and “the application of deep learning to genomic medicine is off to a promising start; it could impact diagnostics, intensive care, pharmaceuticals and insurance, and bridge the ’genotype-phenotype divide’ (Brendan Frey). This class of developments is an important motivation for the OERCompBioMed initiative and the design of the “CBM101x” course aimed for biomedical students at the Master’s (and PhD) level.
Gap in education
Earlier curricula for biomedical students have introduced students with supervised learning techniques that rely on hand-crafted features and require sophisticated design. Deep neural netwoks can automatically learn multilevel hierarchies of features that are invariant to irrelevant variations of samples and thus best suited to be applied on biomedical data that is often noisy. Our approach will be to combine e-learning, next generation data analysis, and biomedical image and omics data into a flexible modular course format. The selection of data and problems stem from two of the most important technologies (imaging and omics) driving biomedical research of today and personalized medicine of tomorrow. The integration of these data will be increasingly important to a larger and larger group of students and researchers in the participating institutions, and beyond. Blending of computational science and biomedicine into a single introductory course, being freely available on a modern Open edX platform to a wide audience of students, researchers and interested laymen, is both innovative and transferable. The idea of a joint transdisciplinary effort among several academic institutions on upcoming and penetrating computational methods and technologies will likely have great impact on future biomedical education, medical research, predictive and personalized medicine, and health care.
The intended learning outcomes are regarding:
- mathematical and statistical modelling techniques in biomedical and clinical applications with examples related to in vivo imaging and integrated quantitative physiology, imaging-derived biomarkers, omics data, and systems biology.
- operational principles of selected measurement devices in biomedical research and clinical practise – from DNA sequencing to MRI scanners.
- the concepts of “big data”, “data analytics”, “machine learning”, and “deep neural networks” with examples from biomedical research and personalized medicine.
- the concepts and importance of “open science”, “data sharing”, and “reproducible research”.
- examples of applications of machine learning and deep learning systems to model biomedical data, and their relevance to future biomedical research and clinical practice.
- importance of mathematical models and computations in the analysis
and understanding of complex molecular, cellular and physiological systems, and disease processes, and the need of cross-disciplinary collaborations in future biomedicine.
- acknowledging the value of “open science”, “data sharing”, and “reproducible research”, and its implementation
- the computational mindset in future biomedicine and clinical practice – pros et cons
- making a quantitative link between images / omics data and important biological concepts
- principles and tools from numerical programming, data analysis, and scientific computing. introduction to modern tools like R, Python, and Jupyter notebooks, and “the cloud” for data storage and computations.
- selecting, installing and running modern software tools for data analysis, visualization & reporting, graphics & video production and communication, relevant to biomedical research
- writing scripts and adapting existing programs for own research project and computational environment
- reading, discussing, and making use of literature in computational biomedicine, modern data analysis and machine learning
- communicating effectively with experts in bioinformatics, imaging, mathematical modelling, and computational science regarding own research challenges
- being able to critically discuss and convey the opportunities and pitfalls of computational models to peers and non-experts.
The “CBM101x” course will be divided into the following modules (Mod):
Mod 1 ”Computational methods and tools” (5 ECTS)
• basic tools for data analysis (pandas, numpy, scipy, scikit-image, scikit-learn), modelling (keras, theano, tensorFlow), scripting, programming (Python), data sharing, and open source (GitHub)
Mod 2 ”Machine learning for omics data” (5 ECTS)
• accessing data repositories for genomewide measurement data across disease, identifying candidate genes and pathways involved in disease pathogenesis, mathematical basis of dimensionality reduction, signal feature analysis and inference, introduction to systems biology
->students are encouraged to include modules from IO2 on basics in disease pathology
Mod 3 ”Image analysis in live cell imaging and in histopathology” (5 ECTS)
• automated characterization of cell size and shape, motion tracking, generating automated workflows, cell segmentation using deep learning and CNN
->students are encouraged to include modules from IO2
Mod 4 ”Quantitative structural and functional MR image analysis in human subjects and animal models” (5 ECTS)
• multimodal MR imaging (3D T1, DWI, PWI, BOLD fMRI), image registration and motion correction; image segmentation (incl. deep convolution neural networks), pharmacokinetic modelling and physiological parameter estimation from dynamic MRI, imaging-derived biomarkers and machine learning.
The content of Mod 1 is a basis for Mod 2, 3, and 4. The latter three can be taken independently of each other.
Assessment and approach
• Two assignments with peer assessment: (i) [Mod 1 only] an assignment related to modern e-infrastructure of computational science (computing environments / IDE, data repositories, source code versioning systems etc.) relevant to both omics data and image analysis, (ii) [each of Mod 2,3,4] an assignment being related to a specific topic within computational biomedicine among several predefined choices. This project will be presented orally for peer assessors enrolled in the course, as a “webinar”, at the end of the course / module.
• Digital final MCQ exam / assessment for each module.
Roles of participating institutions
University of Bergen (UIB) [esp. Mod 1,3,4], University of Turku (UTU) [esp. Mod 1,3], University of Eastern Finland (UEF) [esp. Mod 1,2], and University of Southern Denmark (SDU) [esp. Mod 1,2] will join in the development of the online course. Regarding course content and learning activities the Department of Biomedicine, UiB will be responsible. The Open edX platform Akademix.no will be used to provide the online course as a SPOC/MOOC. Analyses of the quality and usefulness of the developed course with suggestions of improvements will be in collaboration with the Centre for the Science of Learning and Technology (SLATE), University of Bergen.
- Prepare for a common infrastructure (AkademiX) for all the modules (Fall 2017)
- Existing course material at all participating institutions are identified and modified to fit each of modules 1-4. Web-site & the AkademiX entry point for the course is established (Spring 2018)
- Prototype of Mod 1 is developed, running on AkademiX, and being tested on selected students and at the Summer School (Fall 2018)
- Prototypes of Mod 2-4 are developed (interface with IO1- IO3), running on AkademiX, and being tested on selected student groups (Spring 2019)
- Prototypes of all modules are refined according to assessments and available as SPOCs to groups of students at each of the participating institutions and at the Summer School (Fall 2019)
- All modules (as SPOCs) are assessed and modified to be the full CBM101x course available as an on-demand MOOC (Fall 2020).