Jens Schwarzbach studied psychology at the Philipps University in Marburg. From 1994 to 1999 he was a research assistant at the TU Braunschweig. In 1999 he did his doctorate at the TU Braunschwei (summa cum laude). The doctoral thesis was entitled “Priming of eye movements with masked stimuli”. From 1999 to 2001, Jens Schwarzbach was a postdoctoral fellow at the Zanvyl Kriefer Mind / Brain Institute at the Johns Hopkins University, Baltimore, MD, U.S.A. From 2001 to 2005 he was an assistant professor at Maastricht University, the Netherlands.
From 2002-2005, Schwarzbach was Principal Investigator of the “Perception and Attention” working froup at F.C. Donders Center of Cognitive Neuroimaging and from 2006-2015 Associate Professor at the Center of Mind / Brain Science at the Università degli Studi di Trento, Italy. He was also co-director of the magnetic resonance imaging laboratory (2006-2013). In 2015 he was appointed W2 professor for biomedical imaging at the University of Regensburg.
Head of the working group "Biomedical Imaging"
i am head of Biomedical Imaging Group at the Chair of Psychiatry and Psychotherapy.
The focus of my events is in the field of clinical and cognitive neurosciences. We deal with how to use imaging techniques (esoecialy functional and structural magnetic resonance imaging (MRI)) to mark biomarkers for mental illnes, e.g. Anxiety or depression, can develop, which we then want to use in diagnostics, prognosis, monitoring and therapy itself. These events are aimed at a wide range of students in medicine, psychology, biology, physics, as well as bioinformatics and medical informatics.
At the moment, all of my courses are related to AI. The seminar “Machine Learning for Neuroscience”, which I teach with Prof. Lingnau (Chair for Cognitice Neuroscience at the Institute of Psychology), gives an introduction to scientific research and modeling of cognitive processes such as visual perception and object recognition as well as psychological disorders such as anxiety and depression. For example, we are discussing to what extent representations of the so-called “hidden layers” in deeo-learning models correspond to the neuronal representations in the human brain and how changed representations in the brain of mentally ill people could provide us with valuavle information when building models.
My colleagues and I go very different ways here, which we tailor to the prior knowledge of our students. For example Prof. Lingnau and I chose the Problem Based Learning (PBL) format for “Machine Learning for Neuroscience”. We start with a problem statement like “Do androids dream of electic sheep?” or “Deep neural networks work in the same way as the visual cortex” and the students explore the topic in a structured discussion in which they find out what you know, but above all what relevant questions the know too little about. They develop these points independently until the next appointment, at which these topics are presented and summarized at the beginning, before moving on to the next problem statement. The advantage of this method is that students with different prior knowledge bring each other to the stand where discussion is fun.
In the continuation event, which I teach not only because of the circumstances, but out of conviction in digital format, I rely on stron initiative in self-study using freely available online material sich as the Google crash course machine learning, and commercial courses which e.g. the company NVIDIA makes available to cooperating universities. Here students can learn at the level that gets them the most. In weekly video conferences, we rework the critical points and jointly develop project ideas that the students implement in the second half of the semester.
Communication: Students learn to speak in an understandable language about the relationship between AI and machine learning with their department. This will help them to implement their own ideas in the future, too, if they are to lead teams of experts.
Abstraction: MAchine learning, especially deep learning, is currently dominated by object recognition and classification. In my courses, the students should learn to apply this knowledge to completely new questions and understand the essence of machine learning.
Where should I start? I expect AI to be extremely useful in my core area, Precision Medicine and Precision Psychiatry. If we manage to find biomarkers for mental illnesses with machine learning, with which we can e.g. save lives by predicting which therapy will best help a particular depression patient.
Data collection and data protection play a major role in current and future debates. This worries many people. I want students to recognize the potential of AI, even its potential to contain the dangers associated with it. For example, MRI scans for large data collections can be anonymized very well with so-called GANs (generative adversial networks).
I am currently dealing with the question of how one can decode the current emotional state of a test person from brain activation. From this I want to develop a neurofeedback procedure that can possibly be used to treat anxiety and depression.
in addition to above Project for neurofeedback I have dealt with machine learning in MR imaging, e.g. using support vector machines (SVM), we have shown that abstract representations of sensory secisions in the parietal lobe exist in humans, or that fear-conditioned stimuli are confused with phobic stimlui in certain areas of the brain in phobics, here thest subjects with spider fear. This can become important for imaging-based psychotherapy.
I studied psychology with a minor in computer science in Marburg. In the second semester I came across a course on artificial neural networks. Since then, the topic has never left me. In my diploma thesis I simulated visual neglect, a neurological disorder, with neural networks. In my doctoral thesis I used, among other things, neural network simulations to explain the motor processing of unconscious stimuli. During my study and dissertation period, I attended international summer and winter schools to continue my education in this subject area and did an internship in he Neural Network Research Group at BOEING in Seattle.
It#s easy for me. For me, the focus was always on understanding biological systems through artificial systems, or making artificial systems more efficient with kowledge from biology. for a few years now I have been using this knowledge in biomedical imaging to further develop Precision Medicine. The topic inspires me and I want to convey it.
That AI is not artificial for me, but a scientific method with which I try to understand people better. Gaining knowledge is my top priority.
I see all of my work here in Regensburg as interdisciplinary, because I have the interesting outsider role of being a non-medical professor in medicine. You can simly learn a lot, both professionally and in communication. We are currently working with physicists and doctors in nucear medicine on a project idea of how to increase the image resolution of positron emission tomographs (PET) and MRIs with neural networks. Together with colleagues from psychology, we are working on the research of fear and enxiety as well as the basics of cognitive neurosciences and I am looking forward to the scientific exchange with members of the new Faculty of computer Science.
That Ai research will become part or head of an interdisciplinary project dedicated to the research of basic human skills, health and diseases, including their therapies. One of the most important prerequisites happened when the Faculty of Computer Science was founded and with the state government’s AI initiative. Now we are asked to make something of it. My wish regarding the biggest change to the status quo is that Regensburg plays an important role when AI and medicine improve the lives of patients togehter.
It is a great initiative. I like to be there with lectures (e.g. at the DataScienc meetup in February 2020), as well as interdisciplinary courses, internships and scientific collaborations.
I think I’ve written enough for today. Thanks again for your initiative.
Mr. Schwarzbach, thank you for taking the time to answer these questions. We wish you a nice day!