Explainable AI-Driven Neural Activity Analysis in Parkinsonian Rats under Electrical Stimulation
summary
The gist
Explainable AI-Driven Neural Activity Analysis in Parkinsonian Rats under Electrical Stimulation is an investigation into applying Explainable AI techniques to analyze electrocorticogram signals from…
In short
Researchers used explainable AI (XAI) to analyze electrocorticogram (ECoG) signals from Parkinsonian rats receiving electrical stimulation. They found that XAI methods, specifically Layer-wise Relevance Propagation (LRP), identified frequency bands and specific brain regions as key predictors of pre- versus post-stimulation states. This shows XAI can uncover features contributing to neural classification without needing manual expert feature extraction.
Key concepts
- Explainable AI (XAI)
- XAI is a set of techniques used to make the decisions of complex machine learning models understandable to humans. In this study, LRP was used to look inside the EEGNet model and see exactly which parts of the brain activity—like specific frequencies or areas—the AI focused on when deciding if the animal was pre- or post-stimulation.
- Layer-wise Relevance Propagation (LRP)
- LRP is an XAI method that traces how much each input feature contributes to the final output prediction. It works backward from the model's decision to determine which specific neural channels and frequency bands were most important for the classification, offering a transparent view of the AI's reasoning.
- Electrocorticogram (ECoG)
- ECoG signals are recordings of electrical activity directly from the surface of the brain, placed on electrodes. In this research, these signals were recorded before and after electrical stimulation in Parkinsonian rats to study how the brain activity changes in response to treatment.
- Pre- and Post-Stimulation Classification
- This refers to using AI to distinguish between two different states of the rats: one state before receiving electrical stimulation and another state after it. The AI model was trained to correctly identify which state the ECoG signal belongs to, helping researchers understand the neural changes caused by the treatment.
Terminology used across episodes
This episode discusses
- Explainable AI-Driven Neural Activity Analysis in Parkinsonian Rats under Electrical Stimulation · Paper Radio
The paper
Explainable AI-Driven Neural Activity Analysis in Parkinsonian Rats under Electrical Stimulation · Read on arXiv
Department of Computer Science and Engineering, Incheon National University · Department of Nanobioengineering, Incheon National University · Center for Brain-Machine Interface, Incheon National University · Department of Electronics Engineering, Incheon National University
Parkinson's disease (PD) is characterized by motor dysfunction and abnormal neural oscillations. This study combined graphene-based cortical stimulation with an explainable artificial intelligence (XAI) framework to investigate spatial and frequency-related features contributing to pre- and post-stimulation classification in Parkinsonian rats. Electrocorticogram (ECoG) signals were recorded before and after stimulation and classified using EEGNet and DeepConvNet. Layer-wise relevance propagation (LRP) and DeepLIFT were applied to identify frequency bands and recording channels contributing to the classification. The anatomical locations of these channels were mapped onto a cortical map. XAI analyses were performed separately for each animal to examine individual attribution patterns. Our XAI analysis revealed frequency-band- and region-specific contributions to pre- and post-stimulation classification in Parkinsonian rats. Our approach provided complementary insights into the features contributing to pre- and post-stimulation ECoG classification alongside PSD-based analyses. These findings highlight the potential of combining graphene-based cortical stimulation with explainable deep learning to identify classification-relevant features across frequency bands and cortical areas in pre- and post-stimulation recordings.
Transcript
Introduction to the show: ident: Genomics Radio. Generated commentary on the latest computational biology and genomics papers.
Ines: I'm Ines.
Marcus: And I'm Marcus, here with Yuki.
Ines: We're looking at the paper: "Explainable AI-Driven Neural Activity Analysis in Parkinsonian Rats under Electrical Stimulation".
Marcus: Here's the gist of the paper. Explainable AI-Driven Neural Activity Analysis in Parkinsonian Rats under Electrical Stimulation is an investigation into applying Explainable AI techniques to analyze electrocorticogram signals from rats with Parkinson's disease receiving electrical stimulation to uncover key neural activity patterns without relying on expert-driven feature extraction.
Ines: This study demonstrates the potential of XAI in providing unbiased insights into PD-related neural dynamics, suggesting a promising path for real-time monitoring and therapy Parkinson's disease is a neurodegenerative disorder characterized by motor dysfunction and abnormal neural oscillations Traditional neural activity analysis in PD has typically relied on statistical methods, which often introduce bias owing to the need for expert-driven feature extraction.
Marcus: The authors explore an explainable artificial intelligence (XAI) approach to analyze neural activity in Parkinsonian rats receiving electrical stimulation Electrocorticogram (ECoG) signals were collected before and after electrical stimulation using graphene-based electrodes that enable less-invasive monitoring and stimulation in PD.
Ines: The study aims to address existing gaps by exploring the feasibility of applying XAI methods to analyze ECoG signals acquired using graphene-based electrodes in a Parkinsonian rat model.
Paper summary: Ines: So wrapping up the paper "Explainable AI-Driven Neural Activity Analysis in Parkinsonian Rats under Electrical Stimulation," the core thesis is that XAI methods, specifically layer-wise relevance propagation, can move beyond just reporting statistical changes to actually identifying which neural inputs are driving a classification decision.
Marcus: The authors used ECoG signals from PD rats with electrical stimulation and classified them using EEGNet, and then they used LRP to show that specific frequency bands like beta and gamma, in particular, were the key drivers of the model's output.
Yuki: What this means for us is that we can start moving away from just looking at overall power changes across all electrodes and start asking where those specific high-activity areas are located on the cortex.
Ines: It suggests that even when using less invasive stimulation methods like graphene electrodes, the underlying neural mechanisms contributing to the state change are spatially and frequency-specific.
Marcus: They found that this XAI framework offers a way to get unbiased insights into PD-related neural dynamics without having to handcraft every single feature manually before feeding it into the model.
Yuki: This opens up a path for real-time monitoring where you could use those relevance scores to autonomously determine the optimal stimulation timing, which would be really useful for therapy.
Ines: The paper also flags that their method doesn't account for every single potential confounding factor, and they noted that relying on a single XAI algorithm might limit the generalizability of these specific attribution patterns.
Marcus: So, while the results are compelling in terms of identifying region-specific and frequency-specific contributions, future work will need to address those limitations to make this framework more robust across different subjects.
Conclusion: Ines: So, to wrap up this session, we're talking about the paper "Explainable AI-Driven Neural Activity Analysis in Parkinsonian Rats under Electrical Stimulation."
Marcus: Basically, they took ECoG signals from these PD rats after electrical stimulation and used an explainable AI method to figure out what neural activity was driving the classification of pre versus post-stimulation states.
Yuki: And the main point is that this approach lets them find specific frequency bands and brain regions that matter, instead of just looking at overall power changes.
Ines: Right, so it's about mapping those high-relevance features back to where they are on the cortex and what frequencies they belong to.
Marcus: What I'm seeing is that this gives us a way to get insights into the underlying biology without needing every single feature manually picked out before we feed it into the model.
Yuki: For me, it means we can connect these specific neural patterns back to the broader history of how these oscillations change in Parkinson's across different animal models.
Ines: It shows that even with less invasive monitoring like graphene electrodes, there are very distinct neural signatures tied to the disease progression or treatment response.
Marcus: And this framework could actually help us design systems for real-time monitoring, where we could use those relevance scores to decide if stimulation timing needs adjusting on the fly.
Yuki: That potential for autonomous timing is significant because it moves us closer to personalized interventions rather than just standard dosing protocols.
Ines: Of course, the authors were careful about that, though; they mentioned this specific analysis relies on one particular explainable AI technique and a limited number of animals.
Marcus: Yeah, so while the findings are strong in terms of identifying those frequency and region-specific contributions, they're flagging that we need more studies to prove how reliable these attribution patterns are across different groups.
Yuki: So the question for us is how we can build on this to make it a more general tool that works across many different PD patients instead of just one rat cohort.
More episodes
- 2610.33881-Sensorimotor behavior from an untrained brain-and-nerve-cord connectome
- 2610.40344-Morphological Features of Murine Pulmonary Arterial Networks for Diagnosing Vascular Disease
- Daily Summary for 2026-11-19
- 2609.10193-Are You Learning Biological Signal or Shortcuts? Auditing and Mitigating Bias in Protein-Protein Interaction Datasets
- 2610.38236-Generative discovery of chemical reaction networks
- 2507.11176-An Interpretable AI framework Quantifying Traditional Chinese Medicine Principles Towards Enhancing and Integrating with Modern Biomedicine
- The day in conversation for 2026-11-19
- Daily Summary for 2026-11-19
- 2504.03590-Towards a Unified Framework for Determining Conformational Ensembles of Disordered Proteins
- 2603.15217-A multiscale discrete-to-continuum framework for structured population models