Search Abstracts | Symposia | Slide Sessions | Poster Sessions
A Systematic Review of Neurobiological Models of Language Processing: Implications for Enhancing Neurosurgical Precision and Planning
Poster Session B, Friday, October 25, 10:00 - 11:30 am, Great Hall 3 and 4
Akshita Garg1, Venu Balasubramanian1; 1Seton Hall University- CNARL
This review systematically examines the evolution of biological language models from classical localizationist frameworks to contemporary network-based approaches, elucidating their significance in enhancing surgical precision and preserving language functions in patients undergoing brain surgery. The objectives include categorizing language models based on underlying theoretical frameworks and methodological paradigms, analyzing their strengths, limitations, and clinical relevance for neurosurgical planning and execution, and synthesizing existing knowledge to offer novel insights into the neural architectures subserving language processing. Method. An extensive PubMed literature search utilizing queries such as the use of neural models of language processing in neurosurgery, and the implications of language models for neurosurgery yielded numerous qualified articles (total: 126). The seminal Wernicke-Broca model laid the foundation derived from lesion studies and electrical stimulation mapping, though subsequent models challenged this modular view by highlighting the involvement of distributed networks and subcortical structures. Dual-stream models, such as the one originally proposed by Hickok and Popel (2004), which distinguished streams for phonological and semantic aspects, emphasizing dissociation within structural connectivity (Simos et al., 2000; Roux et al., 2015; Sierpowska et al., 2019). Network-based approaches enabled precise mapping of language regions through functional connectivity atlases, while lesion-symptom analyses provided insights into the semantic network's structural connectivity (Langs et al., 2014; Collard et al., 2016; Kuiper et al., 2020; Milton et al., 2021). Intraoperative remapping unveiled cortical plasticity, emphasizing the need for dynamic surgical planning (Ng et al., 2023; Freyschlag & Duffau, 2014; Duffau, 2014; Jin et al., 2021; Rofes et al., 2019; Duffau, 2016; Szalisznyó et al., 2017). Multimodal integration research highlighted the integration of perception and production-based representations for effective speech processing (Cogan et al., 2014; Roux et al., 2015; Almairac et al., 2015; Collard et al., 2016; Nourski et al., 2016; Sierpowska et al., 2017; Ding et al., 2020). Neural tuning models revealed the involvement of a broad perisylvian network in language comprehension (Berezutskaya et al., 2017; Long et al., 2016). Investigations into the subthalamic nucleus challenged the cortical exclusivity view, highlighting subcortical roles in linguistic processes (Whelan et al., 2004; Erlich et al., 2011; Courtemanche et al., 2003; Brown et al., 1997; Haber & Knutson, 2010; Zheng et al., 2019; Zikopoulos & Barbas, 2007; Yin & Knowlton, 2006). By synthesizing diverse models, this review offers a novel perspective emphasizing a multimodal network-based approach accounting for dynamic reorganization of language functions, enabling precise neurosurgical interventions, and optimizing outcomes. Integrating biological language models allows neurosurgeons to achieve precise interventions while preserving critical language functions, improving patient recovery. Future research should focus on developing comprehensive language mapping tools, real-time intraoperative monitoring, personalized rehabilitation strategies, and integration with neuroprosthetics. Potential limitations include data integration complexity, inter-individual variability, limited neuroimaging resolution, and translational challenges. Robust data integration frameworks accounting for individual differences, advanced neuroimaging modalities, and interdisciplinary collaborations are crucial for successful clinical translation.
Topic Areas: Methods, Speech-Language Treatment