Participants will learn about the relationship between a clinical problem and a machine learning problem as well as the influence of training data quality and annotation errors on the ability of algorithms to learn. Participants will be introduced to state-of-the-art AI models and learn how to apply and validate them in practice.
Learning objectives
Once participants have successfully completed this module, you have an understanding for:
- The formulation of a clinical problem as a machine-learning problem The challenges and practicalities of assembling a training set of medical images
- The tradeoffs between manual or automatic labelling of training data How to select a state-of the art model, and have hands-on expe-rience training such a model
- Selecting and applying appropriate metrics for the performance of a medical AI system and have gained hand on experience
- How the results of validating a machine-learning model may affect design choices in a subsequent experiment
Learning content
This module covers a wide range of aspects, such as the introduc-tion to image classifiers, ChexNet dataset, image segmentation and Decathlon datasets, labeling of image data, automatic vs. manual methods, adapting existing code to a new problem: classification
It further contains the modern CNN architectures explained, adap-ting code to a new problem: segmentation, Deep Learning Classi-fiers versus old fashioned modelling, interpreting model perfor-mance, and issues in adapting code, 3D data, multimodal data, modern deep learning
Duration
The completion of this module takes approximately 6 weeks.
Credits
3 ECTS
Fee
The cost of this standalone module is 2'400.– SFr.
Reference
02.004
