Biomarkers – Machine Learning
“Machine learning for diagnosis and prognosis in neuroradiology, from images to -omics” by Jonas Richiardi
Machine learning in radiology: resistance is futile” is the title of a 2019 editorial in the journal Radiology. In this course, we will examine why it is the case, but also temper some of the enthusiasm around the subject – much like level 5 autonomous driving, fully automated diagnosis and prognosis from radiology data is very far from a solved problem. In the first half of the course, we will focus on achievements and challenges using only imaging data, while the second half of the course will open towards the future by highlighting some recent advances in models combining imaging with -omics data, and how Natural Language Processing can be leveraged in clinical environments to enrich image-only modelling.
By Kevin Mader
The diagnosis and treatment of lung cancer has been drastically improved by new imaging methods which generate large number of images where single spots can drastically influence the diagnosis and treatment. For physicians this means a long time must be spent carefully reading images. 4Quant (an ETH Spinoff) together with the University Hospital Basel have demonstrated the potential to radically reduce the reading time without sacrificing quality by using Big Data and Deep Learning approaches. We present the work we have done towards a computer aided staging of Non-Small Cell Lung Cancer (NSCLC).
“Keys for efficient collaborative and reproducible neuroimaging computing research” by Sébastien Tourbier
This lecture will introduce you three topics, which are in the heart for establishing and efficiently using common resources: a common standard for brain imaging data organization, version control systems (for code and data), and software containers. This will help you to become more efficient in your day-to-day neuroimaging computing research tasks as you will learn about the keys to make your work more portable, inter-operable and reproducible.