brainhack PROJECTS

Project #1 / Team Leader: Aneta Lisowska / a.lisowska@sanoscience.org
Persuasive Technology for Digital Behavior Change Intervention
Digital Behaviour Intervention, Wearable, Machine Learning, m-health, Personalisation
Abstract:
Behavior change strategies aim to modify health risk behaviors such as physical inactivity, unhealthy eating and substance abuse to prevent the development of chronic diseases [1] and improve individuals physical and mental well-being [2]. Ubiquitous availability of mobile phones paired with wearable devices provides an opportunity for provision of digital behavior change interventions (DBCI). The effectiveness of DBCI depends on individuals adherence to the intervention. To facilitate adherence, it is important to provide the right support at the right time. The adequate support may include tailored content of notifications and personalised activity suggestions. The right time not only reflects the literal time of the day, but also takes into account the context of the individual (both internal, e.g. emotional state, and external, e.g. location). Machine learning methods can be used both to understand the patient’s internal context (e.g., classify emotions based on a signal captured by the wearable device [3]) and to personalize an intervention (e.g., tailor notification timing [4]). Brain hack participants will have the opportunity to develop methods that facilitate behaviour change. This may include predicting adherence to the activity suggestions, classifying emotional state from consumer-grade wearable1 and finding the conditions under which individuals are responsive to ’nudges’. The solutions to these problems may rely on simple statistical models or on more advance deep learning approaches. A successfully developed solution can be published and potentially even applied in real-life study!

[1]  Dietz, W.H., Brownson, R.C., Douglas, C.E., Dreyzehner, J.J., Goetzel,R.Z., Gortmaker, S.L., Marks, J.S., Merrigan, K.A., Pate, R.R., Powell,L.M., et al.: Chronic disease prevention: Tobacco, physical activity, andnutrition for a healthy start: A vital direction for health and health care.NAM Perspectives (2016)
[2]  Dale, H., Brassington, L., King, K.: The impact of healthy lifestyle inter-ventions  on  mental  health  and  wellbeing:  a  systematic  review.  MentalHealth Review Journal (2014)
[3]  Lisowska, A., Wilk, S., Peleg, M.: Catching patient’s attention at the righttime to help them undergo behavioural change: Stress classification exper-iment from blood volume pulse. In: International Conference on ArtificialIntelligence in Medicine, pp. 72–82 (2021). Springer
[4]  Lisowska, A., Wilk, S., Peleg, M.: From personalized timely notificationto  healthy  habit  formation:  A  feasibility  study  of  reinforcement  learn-ing  approaches  on  synthetic  data.  In:  Proceedings  of  the  AIxIA  2021SMARTERCARE Workshop, CEUR-WS, pp. 7–18 (2021)
[5]  Pinder,  C.,  Vermeulen,  J.,  Cowan,  B.R.,  Beale,  R.:  Digital  behaviourchange  interventions  to  break  and  form  habits.  ACM  Transactions  onComputer-Human Interaction (TOCHI)25(3), 1–66 (2018)
[6]  Eyal,  N.:  Hooked:  How  to  Build  Habit-forming  Products.  Penguin,  ???(2014)
[7]  Kahneman,  D.,  Sibony,  O.,  Sunstein,  C.R.:  Noise:  a  Flaw  in  HumanJudgment. Little, Brown, ??? (2021)
[8]  Sapolsky, R.M.: Behave: The Biology of Humans at Our Best and Worst.Penguin, ??? (2017)
[9]  Fogg,  B.J.:  Tiny  Habits:  the  Small  Changes  that  Change  Everything.Eamon Dolan Books, ??? (2019)
[10]  Shah, R.V., Grennan, G., Zafar-Khan, M., Alim, F., Dey, S., Ramanathan,D.,  Mishra,  J.:  Personalized  machine  learning  of  depressed  mood  usingwearables. Translational Psychiatry11(1), 1–18 (2021)
[11]  Saganowski, S., Kazienko, P., Dziezyc, M., Dutkowiak, A., Polak, A., Dzi-adek,  A.,  Ujma,  M.:  Review  of  consumer  wearables  in  emotion,  stress,meditation,  sleep,  and  activity  detection  and  analysis.  arXiv  preprintarXiv:2005.00093 (2020)

List of materials:
List of requirements for taking part in the project:

1-3 people with bio-med background (including psychology, cognitive science) and knowledge of statistics (Regression, Tests of Significance, ANOVA etc.), 1-3 people with math or computer science background and knowledge of machine learning (SVM, Random Forest, Convolutions Neural Networks, Q-Learning), 1-3 people with physics or engineering background and knowledge of signal denoising approaches. Programming language of choice: Python.

Maximal allowed number of team members: 9
Project #2 / Team Leaders: Alessandro Crimi, Joan Falco Roget / a.crimi@sanoscience.org
Multimodal reservoir causality for effective brain connectivity
Abstract:
The relationship between structure and function is of interest in many research fields involving the study of complex biological processes. In neuroscience in particular, the fusion of structural and functional data can help to understand the underlying principles of the operational networks in the brain. Dynamical causal model and Granger causality have been used in this context to define effective connectivity. Despite the success, those tools have received criticisms as being just predictors of temporal correlation (and not really perturbation based). More recently, new models are emerging from chaos theory and attractors representations. Among those causal representations convergent cross mapping (CCM) is the one receiving a lot of interest in biology and zoology. However, CCM is so far limited to couples of signals/behaviors. In this project we want to investigate this approach for multivariate relationships using recurrent neural networks.

List of materials:
List of requirements for taking part in the project:

Participants should be knowledgeable on Python programming. Signal theory, dynamical system is an asset Neuro anatomy and physiology is welcome.

Maximal allowed number of team members: 8
Project #3 / Team Leaders: Joan Falco Roget, Luca Gherardini / j.roget@sanoscience.org
Network similarities across brain focal lesions
Abstract:
With adequate mathematical and computational methods brain networks can be reconstructed from Magnetic Resonance Images (MRIs) opening the gates for meaningful statistical studies about several brain deseases (e.g. stroke, tumor, alzheimer, ...) In this project we will focus on finding structural similarities between brain networks suffering from stroke and tumors. Possible approaches include network statistics, statistical connectomics and topological data analysis. Next steps would include inferences on common cognitive impairments between the groups.

List of materials:
List of requirements for taking part in the project:

Python
Mathematics
Complex Systems (optional)

Maximal allowed number of team members: 5
Project #4 / Team Leaders: Adam Sobieszek, Hubert Plisiecki / aw.sobieszek@student.uw.edu.pl
Signal Space Generative Adversarial Networks for Modelling EEG Brain Activity and Predicting Emotional Decisions
Abstract:
We recently proposed an architecture for generating EEG signals called a Signal Space Generative Adversarial Network (SigS-GAN), that learns a latent space representation of the signals it was trained on. We impose a regularization on these latent representations of signals, which makes them useful for understanding and predicting the processes that were visible in the EEG activity.

The regularization (which is an extension of Path-Length Regularization to the frequency domain) encourages the learning of a latent space where a distance between two points approximately corresponds to a measure of distance between the two signals that would be generated if we were to put these points into the generator. This is useful as it (a) adds smoothness to the representation, such that signals that are similar correspond to points that are near each other, (b) directions in latent space start to correspond to useful features of the signals, which makes it easier to describe and perform classification, (c) you can use such a latent space in order to perform a new kind of EEG analysis, where you analyze, in the latent space, the differences between point corresponding to signals that, for example, lead to two different decisions.

The goal of this project is to develop the architecture, and create the analysis methods and tool needed to pursue this last opportunity (c) for a new way of EEG analysis. We will brainstorm what modification to the present architecture would make a latent space analysis of EEG signals easier and more fruitful. Implement them, and train the networks on several different datasets of ERP studies, where participants made different types of decisions based on a processing of emotional words. Next we will apply the developed methods in order to explain what differences in electrical activity correlated with different decisions and try to predict them on data unseen by the model. The techniques developed as a part of this project could lead to a scientific publication.

List of materials:
List of requirements for taking part in the project:

- Either knowledge of python (we’ll use PyTorch for the neural net) or mathematics (linear algebra, multi-variate calculus), as we’ll spend some time working out Fourier-analysis-based regularization terms and statistical approaches to the analysis of a trained latent space.

- It is not required to be proficient in the topics discussed in the abstract (GANs, path-length regularization, latent space representations), as we will spend some time at the beginning of the project acquainting ourselves with them.

Maximal allowed number of team members: 10
Project #5 / Team Leaders: Monika Pytlarz, Sylwia Malec / m.pytlarz@sanoscience.org
Renaissance of diffusion models – do they really beat GANs? Stain style transfer and data augmentation for brain histology images.
Abstract:
Assessment of brain tissue can be more precise by combining different histology stainings. Generating digital fluorescence histology would be beneficial, because of fewer artifacts and easier diagnostic recognition than on grayscale images. Staining style transfer is also widely used for the normalization task compensating for variability occurring during the sample preparation. The goal is to implement a transfer style (or data augmentation and transfer style) for generating histology images of different staining. Recent studies describing the renaissance of diffusion probabilistic models suggest that they can prove to be superior to variations of GANs within the transfer style and data augmentation. In the case of researching brain pathologies, we are facing the challenge of lacking publicly available big histology datasets and benchmarks. Data augmentation helps overcome the challenge of small sample size settings, improves the model prediction accuracy, and reduces data overfitting. The aim of the project is to compare these two competitive neural networks – the diffusion model vs GAN – and to determine the winner in the scope of the specified domain.

List of materials:
List of requirements for taking part in the project:

python programming, machine learning, deep learning skills; familiarity with generative adversarial networks, basic knowledge about above mentioned medical imaging modalities may be also useful

Maximal allowed number of team members: 9
Project #6 / Team Leader: Bartek Król-Józaga / kroljoza@agh.edu.pl
Analysis of LC activity during the decision-making process based on changes in pupil size
Abstract:
A high-level cognitive process of decision making (DM) among desirable alternatives requires coordination of distinct cortical and subcortical areas. Computational models can be used to understand these processes, but many of the existing ones focus on simulating only one of the many parallel operations. The existing holistic spiking neural model (https://doi.org/10.5281/zenodo.4280963), which addresses the problem of simulation DM will be our starting point to examine emotional arousal on DM.

The aim of this project will be first to extend the existing model with a population representing the neuromodulatory locus coerelus (LC) functions and then to validate the correlation of its activity with the actual data of changes in pupil size collected using the eyetracker during behavioral test.

Participants will have the opportunity to work in a multidisciplinary team focused on several parallel areas: cloud computing, digital medical signal processing, building a spiking neural network model, and validating cognitive theories regarding decision-making and emotional impact. Our model will land on a supercomputer and we will find out if AI can have emotions!

List of materials:
List of requirements for taking part in the project:

Researcher role: (1-3) people with bio-med background (you will be a substantive support for technicians; your task will be to be able to define the functions of specific areas of the brain in the decision-making process). Knowledge in the field of statistics will also be appreciated.

DSP Engineer: (1-3) your role will be to digitally process the eye tracker signal. Get ready for a task in the field of filtration or implementation of the blink detection algorithm.

Python Dev: (1-3) your task will be to run the existing model on the cloud, implement the spiking neural network populations and give them functions defined by Researchers. Knowledge of the basics of nengo Python library is required.

Programming language of choice: Python!

Maximal allowed number of team members: 9
contact us
Contact us today to take the first step of your
journey with Brainhack Krakow!
© 2022 Jagiellonian University in Kraków