Alzheimer's is often not diagnosed until serious problems with memory and thinking have already occurred. Louise Bloch, a doctoral student at Fachhochschule Dortmund, is relying on machine learning to start potential therapies earlier. The computer scientist wants to train her software so that Alzheimer's can be diagnosed at a very early stage using brain scans, among other things.
"Computers, or more precisely supervised machine learning processes, are very good at recognizing complex patterns," says Louise Bloch. This is why the use of technology in a heterogeneous disease such as Alzheimer's, which is influenced by many different factors, can be of great benefit. At Fachhochschule Dortmund, the doctoral student is feeding her program with data from a total of more than 2,000 test subjects. The machine is designed to recognize patterns in the large number of brain scans, which show different stages of the disease and some of which were taken before an Alzheimer's diagnosis.
Is there a risk of Alzheimer's?
The data model developed by Louise Bloch is currently learning several 100,000 parameters in order to recognize the correlations between MRI images, other clinical data and the stages of the disease: Which changes are relevant? Which brain regions are important? This creates a neural network that can evaluate large amounts of data "If this trained process then receives data from patients with an unknown diagnosis, the patterns can be compared and a diagnosis made," says the computer scientist. Louise Bloch is also convinced that a possible risk of Alzheimer's can then be predicted.
"However, the diagnosis or even a potential risk of Alzheimer's must ultimately be confirmed by a doctor," emphasizes the scientist. Part of her research at Fachhochschule Dortmund is therefore also looking at how the machine's data can be processed in such a way that it can be transparently understood by doctors. Louise Bloch is working closely with the medical faculty at the University of Duisburg-Essen. As is usual at universities of applied sciences, she is doing her doctorate in cooperation with both universities. The UAS doctoral student has just been granted two more years of research for her project and has published the results of her work together with Prof. Dr. Christoph Friedrich from the Faculty of Computer Science in the journal SN Computer Science. The publication was scientifically peer-reviewed.
Publication
Bloch, L., Friedrich, C.M. & for the Alzheimer's Disease Neuroimaging Initiative. Machine Learning Workflow to Explain Black-Box Models for Early Alzheimer's Disease Classification Evaluated for Multiple Datasets. SN COMPUT. SCI. 3, 509 (2022). https://doi.org/10.1007/s42979-022-01371-y(Opens in a new tab)
Explanatory video
This explanatory video (English with German subtitles) describes the research field of Louise Bloch, a doctoral student at the Graduate Center of the Fachhochschule Dortmund. It was produced by her for the DART Symposium at Fachhochschule Dortmund(Opens in a new tab) . Her video won the "Young Science Communication Award" for science communication.
