Posters

 P2. Understanding the role of  cognitive control for repetitive transcranial magnetic stimulation in Depression

Pranjul Verma; Sanket Houde; Hari Prakash Tiwari; Nandini Priyanka B; Alok Bajpai; Pragathi P. Balasubramani

Translational Neuroscience and Technology Lab, Department of Cognitive Science, IIT Kanpur

Abstract:  Depression is a prevalent mental health disorder marked by persistent low mood, anhedonia, and cognitive deficits, often resistant to conventional treatments. Repetitive transcranial magnetic stimulation is a promising intervention for Depression.
However it has only about 30-40% remission rates, demanding personalization and optimization of this method for maximal efficiency.
Through this study, we aim to explore the use of simultaneous EEG and repetitive Transcranial Magnetic Stimulation (rTMS) to understand neural connectivity changes during various cognitive tasks in individuals with MDD. We aim to identify tasks that enhance fronto-parietal connectivity and exhibit anticorrelation between the dorsolateral prefrontal cortex (dlPFC) and subgenual anterior cingulate cortex (sgACC), that is hypothesized to improve depression symptoms. Methods involve 64-channel EEG recording during the administration of rTMS (10 Hz) while participants perform tasks that facilitate resting or interoceptive or external attention or reward processing functions, in a pre-TMS, during TMS, and post-TMS design. This study includes 15 participants, with each session lasting around 150 min.
Our primary outcome measures are the changes in connectivity patterns with reference to our hypothesis, focusing on task based difference and the improvement of depressive symptoms through different self report questionnaires, which are PHQ-9, GAD-7 & MADRS.
Findings from this study will inform a personalized approach, where the task that engages the neural circuitry for maximal improvement of depression symptoms could be used for each participant during rTMS intervention.

 P4. Graph learning methods for functional brain connectivity in naturalistic stimuli fMRI

Sasanka GRS; Ayushi Agrawal; Santosh Nannuru; Kavita Vemuri

International Institute of Information Technology - Hyderabad (IIIT - H)

Abstract:  Functional MRI research using naturalistic stimuli like movies examines how brain networks support complex processes like empathy. Our study proposes a novel fMRI signal processing pipeline with FIR high-pass filtering, voxel-level K-means clustering, and a sparsity-based graph learning approach. Applied to whole-brain timeseries data from 14 healthy volunteers watching two short movies, this method outperforms traditional approaches, achieving an 88% match with the emotion contagion scale, a ground truth for behavioural response to stimuli. Temporal analysis reveals gradual induction of empathy, supported by the method’s ability to capture dynamic connectivity patterns. Connectome analysis highlights the importance of the Insula, Amygdala, and Thalamus in emotional processing, with lateral brain connections promoting synchronized activations. Spectral filtering isolates Amygdala and Insula, further emphasizing their role in emotional and empathetic processing during high emotional states. These insights deepen understanding of the neural dynamics of empathy, paving way for targeted interventions in empathy-related conditions.

 P6. Early prediction of pharmaco-intervention outcome for Depression using gut-brain electrophysiology- a longitudinal study

PAmal Jude Ashwin(1); Hari Prakash Tiwari(1); Alok Bajpai(2); Nandini Priyanka(3); Pragathi P Balasubramani(1)

(1) Translational Neuroscience and Technology Lab, Department of Cognitive Science, IIT Kanpur. (2) Health Center, IIT Kanpur (3) Neuroclinical Innovative Solutions Pvt. Ltd.

Abstract:  

Globally, 970 million people live with a mental disorder. Nearly, 280 million people worldwide suffer from depression. Medication is the first step of treatment against depression. However, 50% are non-responders to the SSRI strategy and treatment requires switching or augmentation with other medications, altogether demanding the need for efficient principles for arriving at a personalized treatment strategy.

In previous literature, predictive models using ML techniques and cross sectional, baseline electrophysiology (EEG) have been used to predict the treatment outcome and guide the choice of treatment plan. However, the reliability (consistency) of a cross sectional recording has been reported to be low and there is a need for longitudinal tracking for reliable prediction. And moreover, while EEG monitors only the neural state of the participant, a whole person approach considering the gut physiology is suggested to be sensitive to depression condition and its treatment. Therefore, monitoring the gut state along with the neural state may enhance the predictability and reliability of treatment outcome.

In our study, we collected EEG and EGG data from 135 participants from three different time points from the start of the antidepressant medication. Our initial analysis suggests that longitudinally tracked, quantitative electrophysiological features significantly enhance the accuracy of treatment outcome prediction, and provides support for early prediction of treatment outcome, thereby guiding the medication strategy.

 P8. Automation of motor assessment using web cam feed

Barath Kumar; Sundari Elango; V Srinivasa Chakravarthy

CNS Lab, Department of Biotechnology, IIT Madras, India

Abstract:  The Fugl-Meyer assessment (FMA) are the most widely used tools for assessing upper extremity motor function in stroke survivors. However, due to its subjective and time-consuming nature, it must be directed by therapists in a hospital or clinic environment, and it is not quite suitable for use at home. In this paper, we propose a deep learning-based system for analyzing various FMA tasks. Using a convolutional neural network architecture (Mediapipe), we extract human pose data and use it to generate the necessary features. The FMA scores are obtained using a feature selection approach combined with a Deep learning models based on the FMA’s linguistic guidelines. This framework could be used to assess upper extremity function during stroke recovery to relieve therapists of the time-consuming, repetitive process. It could also be done at home because it does not require any sensors and only uses a computer with a webcam or even a smart phone.

 P10. Modeling the exploratory dynamics of the Indirect Pathway in the Basal Ganglia using a network of Chaotic Attractors

Palika Charitha; Jayant Sharma; Nurani Rajagopal Rohan; S.S. Nair; V. Srinivasa Chakravarthy

CNS Lab, IIT Madras

Abstract:  Parkinson's Disease (PD) is a neurodegenerative disorder, caused by the loss of dopaminergic neurons in the Substantia Nigra pars compacta (SNc). We had earlier proposed that when the BG is described using Reinforcement Learning (RL), the Direct Pathway subserves exploitation, while the Indirect Pathway subserves exploration. The putative role of the Subthalamic Nucleus (STN) in exploration, is supported by the loss of complex dynamics in STN under PD conditions. We describe a model of the Basal Ganglia (BG) in which loss of complex dynamics in STN, which is modeled by a network of chaotic Rossler systems, is manifested as impaired performance in Iowa Gambling Task (IGT).  The dynamics of the Rossler network are characterised by the parameter ‘a’, which allows the network to move from a periodic to a chaotic regime, characterised by uncorrelated desynchronised oscillations. The network is tuned to exhibit chaotic behaviour in normal conditions and synchronised periodic oscillations in PD conditions. The network receives feedback via mean-field diffusion from SNc, which controls the collective behaviour of the network. This network also receives inputs (Iext) from the striatum (D2R neurons), which can induce STN transition into a periodic regime. The BG circuitry is trained for the Iowa Gambling task (IGT) using RL. Temporal difference error (δ)  is analogous to dopamine in SNc. We define epsilon (ε) as the exploratory parameter, which is a function of δ and linearly controls ‘a’ in STN. The initial stages of training demand a greater exploration (high ε), which necessitates the network to operate in a chaotic regime. To simulate the dopamine loss in PD condition, the δ value is delimited which in turn constrains ε, restricting STN to a periodic regime. This limits the BG circuit's ability to learn the IGT task. 

 P12. Graph wavelet diffusion for brain structure to function mapping

 Chirag Jain (1); Avinash Sharma (2); Upadrasta Naga Sita Sravanthi (1);  Bapi Raju Surampudi (1)

(1) IIIT Hyderabad; (2) IIT Jodhpur

Abstract:  The relationship between brain functional connectivity (FC) and structural connectivity (SC) is examined using models that simulate activity on SC to derive FC through various methodologies, including graph diffusion kernels. However, current studies lack the correlation between diffusion scales and specific brain regions of interest (RoIs), limiting the effectiveness of graph diffusion. We introduce a novel approach using graph wavelet heat diffusion to determine the diffusion scale for each RoI, accurately predicting FC with advanced metrics. Using open HCP data and the Desikan-Killiany atlas (87 RoIs), our method improves FC prediction and extracts significant properties from SC and FC, enhancing the understanding of whole-brain connectivity and highlighting the potential of graph wavelet heat diffusion.

 P14. A Deep Learning Model of Spatial and Object-responsive neurons of Hippocampal Formation.

 Azra Aziz; Bharat Patil ; V. Srinivasa Chakravarthy

Computational Neuroscience Lab, Department of Biotechnology Bhupat and Jyoti Mehta School of Biosciences, IIT Madras

Abstract:  Navigation and memory are vital for survival, with organisms relying on landmarks and objects for orientation and foraging. The hippocampal formation constructs a spatial cognitive map using spatial cells like Place, Grid, and Head Direction cells. Sensory processing guides this map's construction, with the Medial Entorhinal Cortex (MEC) encoding self-motion cues (Path Integration) for the 'where' pathway, while the Lateral Entorhinal Cortex (LEC) integrates sensory inputs for the 'what' pathway. Object-dependent spatial cells, including Object Vector cells, Object Sensitive cells, Border cells, Landmark Vector cells, and Object Trace cells, have been identified, shedding light on how organisms navigate and remember their surroundings. Many computational models address these representations individually or in subsets, but only some comprehensively explain them within a single model. We propose a deep neural network with two separate pipelines for Path Integration and vision. A Graph Convolution Neural Network (GNN) integrates these inputs, followed by a feed-forward network. A virtual agent traverses a 3D environment; its locomotion cues are the Path Integration input, and POV (Point of View) is the vision input for the model. The model's output is the Heading Direction, Reward and Agent's current position. The choice of environment and learning rate for different outputs leads to the emergence of all aforementioned spatial cells. This comprehensive modelling approach is novel as it propagates a simplistic theory of multisensory integration using GNN, explains all observed spatial representations and can be extended as a complete navigation model using a reinforcement learning agent in future.

 P16. Towards building a real-time seizure prediction and classification system 

Lazar Tony; Anshuman Sahoo; Rathore BP; Nandini Priyanka B; Pragathi P Balasubramani

IIT Kanpur

Abstract:  Epilepsy affects around 50 million people worldwide [1], causing spontaneous seizures. While drugs help many, side effects and treatment resistance are seen in one-third of the population [2]. Even surgery doesn't guarantee seizure freedom [3]. The current epilepsy monitoring system is based on manual EEG analysis that is time-consuming, expensive, prone to errors, and inconsistent between experts, hindering patient care. Though several studies have shown high efficiency in EEG-based techniques for the classification and prediction of seizures on existing databases, there is a need to validate these techniques: (1) using multiple clinical databases and (2) for the continuous monitoring of EEG signals [4]. In our study, we extract features from EEG to create a classification and later a prediction system for seizures that can be deployed in clinics. Our current ongoing work is to assess the utility of our seizure detection system through field-level validation and facilitate its integration with daily life for clinical utility.

References

1.  World Health Organization: WHO. Epilepsy. Published February 7, 2024. https://www.who.int/news-room/fact-sheets/detail/epilepsy

2.  Perucca E, Perucca P, White HS, Wirrell EC. Drug resistance in epilepsy. The Lancet Neurology. 2023;22(8):723-734. doi:10.1016/s1474-4422(23)00151-5

3.  Téllez-Zenteno JF, Dhar R, Wiebe S. Long-term seizure outcomes following epilepsy surgery: a systematic review and meta-analysis. Brain. 2005;128(5):1188-1198. doi:10.1093/brain/awh449

4.  Acharya UR, Sree SV, Swapna G, Martis RJ, Suri JS. Automated EEG analysis of epilepsy: A review. Knowledge-Based Systems. 2013;45:147-165. doi:10.1016/j.knosys.2013.02.014

 P18. Towards evaluating the differences between reward processes and executive control for successful brain machine interfacing recovery in post-Stroke patients with upper limb impairments

Gokul Krishna Raja P; Pranjul Verma; Nikunj Bhagat; Pragathi P. Balasubramani

Translational Neuroscience and Technology Lab, Department of Cognitive Science, IIT Kanpur

Abstract:  Stroke is one of the leading causes of disability and death worldwide. Advanced technologies such as Brain Machine/Computer Interfaces (BMI/BCI) can improve the efficacy of rehabilitation therapies through a multidisciplinary approach using robotic exoskeletons for motor recovery. However, the motor recovery for a BMI algorithm seldom looks into the cognitive aspects that are conceptually thought as principal components projecting the motor performance. In such interfaces, exploring  the network level understanding of the source space, in terms of various cognitive dimensions such as executive control versus reward processing is fruitful in both improving the therapy protocols further and understanding the subject level differences. We are analysing the data from 18 post-stroke impaired patients where each subject had to undergo 12 sessions of an EEG based BMI task for promoting hand recovery. In the task, participants had to try initiating ‘a motor intent’ which would be detected by an EEG classifier and further validated using the EMG signals, and successfully start the exoskeleton to reach a cue displayed on the screen. Using the Recursive Sparse Bayesian Learning algorithm, we are looking into the source level activations before and after both the successful and unsuccessful markers. We use an Actor-Critic architecture that incorporates a framework of critic (value, Cingulo opercular Network) and actor (executive policy, Fronto parietal Networks) in order to understand the differential contributions mediating processing of reward, attention, and executive control. These understandings will have a huge importance to efficiently improve rehabilitation therapies by appropriately accounting for the cognitive measures to adapt the motor BMI and improving its efficiency, together bridging the gap between cognition and behaviour in BMI.

 P20. The Individualized Music Program for the people living with Dementia: a feasibility study for Indian population

Arijit Bhattacharya(1,2); Faheem Arshad (2); S Sandeep Kumar (2); Suvarna Alladi (2)

(1) Department of Psychology, Ashoka University, India; (2) Department of Neurology, National Institute of Mental Health and Neurosciences, India

Abstract:  Recent advancement in individualized music-based interventions has demonstrated to have positive effects on the overall quality of life in persons living with dementia. Likewise, music-based intervention in India has made great progress in recent years whereby it remains of crucial importance to understand the feasibility and short- and long-term benefits of the individualized music and memory program specifically for the people living with dementia in India. We adapted and modified of the MUSIC & MEMORY® questionnaires and recruited clinically diagnosed dementia patients followed by cognitive assessments (n = 14). Further we prepared personalized playlists for individual patients. A weekly and a final 3 month follow-up was conducted to monitor the progress (n=7). The result showed that the patients are spending their leisure time listening to music suggested in the playlist. Most of the patients showed reduced agitation and restlessness while listening to the music, and that they feel less sad or bored. Although the study is still on going but the reports from the caregiver’s and patients suggests that the individualized music program is feasible and useful in Indian populations with dementia.

 P22. Studying the spatial disorientation in mild cognitive impaired patients using a path integration model and electrophysiology

Adrija, Mansimran; Hari Tiwari; Nandini Priyanka; Subasree Ramakrishnan; Avinash Singh; Pragathi Balasubramani

Indian Institute of Technology, Kanpur

Abstract: Spatial disorientation represents an early indicator of Alzheimer Disease*. Mental overload is known to induce specific path integration shortfalls*. Errors in spatial cognitive mapping may be induced by stress, mental overload, or reduced attention capacity. These factors interfere with perceiving the dynamic environment and path integration elements. Obstacle-filled paths hinder effective path mapping between home and destination.

We aim to probe these stress/attention factors to understand the affect on spatial navigation and path integration in context of MCI patients. We employ a multifaceted approach that combines real-set up of spatial navigation tasks with continuous interruption with a working memory overload. Mismatched negativity paradigm is implemented with non invasive EEG, on 65 year + (N=8) individuals screened as normal and MCI through MMSE. Behavioral and EEG parameters used as measures.

Initial results : Different cluster of behaviors can be observed across trials based on working task and spatial task performances : 1) increasing deterioration in navigation with increasing load and resource constraints 2) Resource allocation only to navigation, without heed to memory load 3) Simultaneously and adequately performing both tasks. We study spatial cognition and their neural basis and understand the mechanisms behind these clusters of behaviors through path integration model.