The upsurge of Machine Learning and Deep Learning algorithms paired with the high amount of digital data generated in radiology is changing this medical specialty. Deep Learning algorithms, in particular convolutional networks, are promising techniques for processing medical imaging data in a wide variety of clinical scenarios.
The first session will be focused on data acquisition and manipulation in order to optimize for a Deep Learning task, what the main hindrances are and how to solve them. We will also cover the basic requirements and tools needed for DL in the radiology setting.
The second session will cover a hands-on neural network training pipeline in which we will tweak the parameters of a neural network to understand how they influence the final result. We will also provide tips and advice on how to start the DL journey in your own department.
The last session will be dedicated to model evaluation, how the main metrics are useful in different case scenarios, and their limitations in a “real-life” clinical setting. We will also review and work on the guidelines for critical assessment of AI research.
Throughout the course we will also see other applications of DL in the healthcare setting, such as study protocolization, structured reporting and communication of incidental findings, among others.
We will also cover the main limitations of Deep Learning in the field of medical imaging and discuss basic notions and resources to start experimenting with Deep Learning algorithms in the daily clinical practice. Each day, a series of selected topics will be presented, followed by a guided interactive activity and group discussion. At the end of every session, we will take some time to explain and debate Deep Learning implementation scenarios in the field of Radiology.
Maximum number of seats: 10
- Learn in a practical way how a neural network works, its strengths and limitations.
- Familiarize yourself with key concepts of DL evaluation to be able to critically appraise the scientific literature.
- Discover resources and tools to start experimenting with DL models in your daily practice.
Programme will include
- Lectures on key points of the Deep Learning pipeline: data preparation, model training and model evaluation.
- Individual guided “hands-on” sessions with a real convolutional neural network, with group discussion of the results.
- Lectures on critical reading, as well as on opportunities and obstacles of the practical implementation of Deep Learning tools in the daily workflow of a Radiology Department.
Level II - General radiologist
- A computer with Google Chrome and Internet Access