New artificial intelligence research should optimize the entire area for scans. The research is expected to create benefits for both the patients and the staff. Next step in the research is to look into the areas with limited training data. DTU Professor Anders Bjorholm Dahl will give you insights into the latest research at Medico Bazar 2020.
He has answered a number of questions prior to his presentation at the Point of View Session about how to use Artificial Intelligence (AI) for imaging technologies in hospitals.
What kind of research do you do in the health domain?
My research is in methods to quantify structures in image, especially 3D images. Imaging is widely used in clinical practice and in preclinical research, and the analysis methods that my group and I develop are, to a wide extent, applied to such data.
We focus on methods that allow detecting relevant structures and measuring their size and shape. Often, such measures are not well defined, so another core element of our research is to find mathematical descriptors that relate to a given use-case, says Professor Anders Bjorholm Dahl.
Which kind of potential does AI for image analysis have?
The basic principle in AI for imaging is mathematical models that can translate image data from pixels to information that can either support our decision making or directly make decisions for us.
With AI, we aim at doing this without explicitly modeling the relation between the input image data and the information, but instead have generic models where we can automatically adjust their parameters based on training data.
That is data where we have images and the associated information. This will typically make the modeling easier and allow for applying such models in all areas where relevant training data is available, says Professor Dahl.
What are the next steps?
In many cases, AI methods and especially deep learning models have very high performance, and problems that previously did not have a solution, now we have one. Still the methods require much training data, and not all problems are easy to solve with AI, so there is a need for research in problems with limited training data.
Also, despite the high performance and therefore high reliability of the methods, it can be difficult to tell when deep learning methods fail or when their prediction is uncertain. This is another area that needs further research. It is also important that AI is used in an ethical way, which also requires research, says Professor Dahl.
Why is AI and image research important in the health care sector?
AI is important because it has the potential for supporting diagnosis and decisions made from image data and take over trivial manual tasks, which can free up time and improve quality in the health sector, says Professor Dahl.
Who would you like to meet at Medico Bazar 2020?
It would be interesting to discuss the potentials of AI and image analysis within the health care sector with regional decision makers, clinicians, companies, computer scientists and research foundations.
At the other sessions you can hear about:
- How Cyber Security challenges in the health care sector concern everyone
- How biological data can be crunched to create better medical practices
- How to implement new technologies at hospitals