Biology subjects

Mahmud, M. S.

Publications and source records attributed to Mahmud, M. S..

5 recordsLinked to original sources

bioRxiv2026

Neural microstates underlying categorical speech perception using Bayesian nonparametrics

Categorical perception (CP) reflects the human auditory systems ability to map continuous acoustic signals onto discrete categories. Understanding the relationship between neural dynamics and perceptual decisions is central to speech-language processing. Here, we implemented a data-driven approach using Bayesian nonparametrics and machine learning to characterize the relationship between auditory cortical responses and speech categorization behaviors. By applying a hierarchical Dirichlet process hidden Markov Model (HDP-HMM) to source-reconstructed event-related potential (ERP) data, we identified temporally distinct neural microstates that capture the evolving stages of speech categorization without imposing predefined temporal windows. Machine learning classifiers, including extreme gradient boosting (XGBoost), support vector machines, and random forests, were applied to decode prototypical (Tk1/5) versus ambiguous (Tk3) speech sound tokens from the microstate data. Using whole-brain activity, the XGBoost classifier achieved the highest decoding accuracy of 94.1% with an area under the curve (AUC) 94.1% within a specific encoding microstate occurring approximately 200-250 ms after stimulus onset. A reduced set of 15 informative brain regions identified via Shapley additive explanations (SHAP) yielded comparable classification performance (90.3% accuracy; AUC 90.0%), with many regions localized to frontal, temporal, and parietal regions in the left hemisphere. Furthermore, neural activity within these regions robustly predicted listeners behavioral identification slopes (R2 = 0.92, p < 0.00001), linking microstate-specific cortical dynamics to individual differences in perceptual gradiency of speech perception. These findings demonstrate that speech categorization emerges within temporally discrete neural microstates during early sensory-perceptual encoding and is supported by a selective, distributed cortical network with clear behavioral relevance.

neuroscience↗

bioRxiv2021

Vestibular Agnosia is linked to abnormal functional brain networks

Vestibular Agnosia - where peripheral vestibular activation triggers the usual reflex nystagmus response but with attenuated or no self-motion perception - is found in brain disease with disrupted cortical network functioning, e.g. traumatic brain injury (TBI) or neurodegeneration (Parkinsons Disease). Patients with acute focal hemispheric lesions (e.g. stroke) do not manifest vestibular agnosia. Thus brain network mapping techniques, e.g. resting state functional MRI (rsfMRI), are needed to interrogate functional brain networks mediating vestibular agnosia. Whole-brain rsfMRI was acquired from 39 prospectively recruited acute TBI patients with preserved peripheral vestibular function, along with self-motion perceptual thresholds during passive yaw rotations in the dark. Following quality-control checks, 25 patient scans were analyzed. TBI patients were classified as having vestibular agnosia (n = 11) or not (n = 14) via laboratory testing of self-motion perception. Using independent component analysis, we found altered functional connectivity in the right superior longitudinal fasciculus and left rostral prefrontal cortex in vestibular agnosia. Moreover, regions of interest analyses showed both inter-hemispheric and intra-hemispheric network disruption in vestibular agnosia. In conclusion, our results show that vestibular agnosia is mediated by bilateral anterior and posterior network dysfunction and reveal the distributed brain mechanisms mediating vestibular self-motion perception.

neuroscience↗

bioRxiv2020

Speech categorization is better described by induced rather than evoked neural activity

Categorical perception (CP) describes how the human brain categorizes speech despite inherent acoustic variability. We examined neural correlates of CP in both evoked and induced EEG activity to evaluate which mode best describes the process of speech categorization. Using source reconstructed EEG, we used band-specific evoked and induced neural activity to build parameter optimized support vector machine (SVMs) model to assess how well listeners speech categorization could be decoded via whole-brain and hemisphere-specific responses. We found whole-brain evoked {beta}-band activity decoded prototypical from ambiguous speech sounds with ~70% accuracy. However, induced {gamma}-band oscillations showed better decoding of speech categories with ~95% accuracy compared to evoked {beta}-band activity (~70% accuracy). Induced high frequency ({gamma}-band) oscillations dominated CP decoding in the left hemisphere, whereas lower frequency ({theta}-band) dominated decoding in the right hemisphere. Moreover, feature selection identified 14 brain regions carrying induced activity and 22 regions of evoked activity that were most salient in describing category-level speech representations. Among the areas and neural regimes explored, we found that induced {gamma}-band modulations were most strongly associated with listeners behavioral CP. Our data suggest that the category-level organization of speech is dominated by relatively high frequency induced brain rhythms.

neuroscience↗

bioRxiv2020

Data-driven machine learning models for decoding speech categorization from evoked brain responses

Categorical perception (CP) of audio is critical to understand how the human brain perceives speech sounds despite widespread variability in acoustic properties. Here, we investigated the spatiotemporal characteristics of auditory neural activity that reflects CP for speech (i.e., differentiates phonetic prototypes from ambiguous speech sounds). We recorded high density EEGs as listeners rapidly classified vowel sounds along an acoustic-phonetic continuum. We used support vector machine (SVM) classifiers and stability selection to determine when and where in the brain CP was best decoded across space and time via source-level analysis of the event related potentials (ERPs). We found that early (120 ms) whole-brain data decoded speech categories (i.e., prototypical vs. ambiguous speech tokens) with 95.16% accuracy [area under the curve (AUC) 95.14%; F1-score 95.00%]. Separate analyses on left hemisphere (LH) and right hemisphere (RH) responses showed that LH decoding was more robust and earlier than RH (89.03% vs. 86.45% accuracy; 140 ms vs. 200 ms). Stability (feature) selection identified 13 regions of interest (ROIs) out of 68 brain regions (including auditory cortex, supramarginal gyrus, and Brocas area) that showed categorical representation during stimulus encoding (0-260 ms). In contrast, 15 ROIs (including fronto-parietal regions, Brocas area, motor cortex) were necessary to describe later decision stages (later 300 ms) of categorization but these areas were highly associated with the strength of listeners categorical hearing (i.e., slope of behavioral identification functions). Our data-driven multivariate models demonstrate that abstract categories emerge surprisingly early ([~]120 ms) in the time course of speech processing and are dominated by engagement of a relatively compact fronto-temporal-parietal brain network.

neuroscience↗

bioRxiv2019

Decoding age-related changes in the spatiotemporal neural processing of speech using machine learning

Speech comprehension in noisy environments depends on complex interactions between sensory and cognitive systems. In older adults, such interactions may be affected, especially in those individuals who have more severe age-related hearing loss. Using a data-driven approach, we assessed the temporal (when in time) and spatial (where in the brain) characteristics of the cortexs speech-evoked response that distinguish older adults with or without mild hearing loss. We used source montage to model scalp-recorded during a phoneme discrimination task conducted under clear and noise-degraded conditions. We applied machine learning analyses (stability selection and control) to choose features of the speech-evoked response that are consistent over a range of model parameters and support vector machine (SVM) classification to investigate the time course and brain regions that segregate groups and speech clarity. Whole-brain data analysis revealed a classification accuracy of 82.03% [area under the curve (AUC)=81.18%; F1-score 82.00%], distinguishing groups within [~]50 ms after speech onset (i.e., as early as the P1 wave).We observed lower accuracy of 78.39% [AUC=78.74%; F1-score=79.00%] and delayed classification performance when the speech token were embedded in noise, with group segregation at 60 ms. Separate analysis using left (LH) and right hemisphere (RH) regions showed that LH speech activity was better at distinguishing hearing groups than activity measured over the RH. Moreover, stability selection analysis identified 13 brain regions (among 1428 total spatiotemporal features from 68 regions) where source activity segregated groups with >80% accuracy (clear speech); whereas 15 regions were critical for noise-degraded speech to achieve a comparable level of group segregation (76% accuracy). Our results identify two core neural networks associated with complex speech perception in older adults and confirm a larger number of neural regions, particularly in RH and frontal lobe, are active when processing degraded speech information.

bioengineering↗