A 3.5-minute-long reading-based fMRI localizer for the language network

Greta Tuckute, Elizabeth Jiachen Lee, Aalok Sathe, Evelina FedorenkobioRxiv 2024 · DOI

TL;DR

Researchers have developed a significantly faster fMRI task (3.5 minutes) to pinpoint language-processing areas in the brain. This "speeded reading" localizer is not only more time-efficient but also more effective than the standard method, showing clearer distinctions in brain activity related to language.

What Problem Does This Paper Address?

Identifying specific brain regions responsible for particular cognitive functions, like language, is crucial for understanding the brain. However, the exact location of these functional areas can vary considerably between individuals. Traditional fMRI studies often average data across many people, which can mask these individual differences and reduce the precision of findings. To address this, researchers use "localizer" tasks – specific experiments designed to reliably activate a target brain region in each individual. The challenge is that these localizer tasks can be time-consuming, which is a significant constraint in fMRI research, especially when studying certain populations or when participants have limited tolerance for the scanner. This paper aims to create a more efficient, yet equally effective, localizer for the fronto-temporal language network.

Methods

The core innovation is a speeded version of a well-established reading-based language localizer. The standard localizer presents words and non-word sequences at a rate of 450 milliseconds per item. The new "speeded" version reduces this to 200 milliseconds per item, a ~56% increase in presentation speed. This rapid serial visual presentation (RSVP) method is inspired by behavioral studies showing that people can process language effectively even at these faster rates.

The study involved 24 neurotypical adults who completed both the standard and speeded versions of the localizer task. Brain activity was measured using fMRI. The researchers then used a group-constrained, subject-specific (GSS) approach to define functional regions of interest (fROIs) for both the language network and the domain-general Multiple Demand (MD) network. This involved correlating activation patterns across runs and comparing the magnitude and selectivity of BOLD responses between the two localizer versions. Statistical analyses primarily employed linear mixed-effects (LME) models to account for individual differences and fROI variability.

Datasets

The study utilized custom-built stimulus materials for the language localizer tasks. The sentences were drawn from the Brown corpus, and non-words were generated using the 'Wuggy' software, designed to adhere to English phonotactic constraints. The materials for the speeded localizer were a new set created to avoid overlap with the standard version and ensure generalizability. No external datasets were used for training or evaluation in this specific study.

Key Insights and Ideas

  • Speeding up is effective: The core idea is that by presenting stimuli much faster, you can achieve the same or better localization results in a fraction of the time. This leverages the brain's capacity for rapid processing.
  • Language network is robust: The speeded localizer reliably identifies the language network, demonstrating its resilience to faster processing demands.
  • Increased selectivity: The speeded localizer not only identifies language areas but does so with a stronger contrast between language-related stimuli (sentences) and control stimuli (non-words). This suggests it's better at isolating the neural signature of language processing.
  • MD network engagement: The speeded reading task, due to increased processing difficulty or attentional demands, engages the domain-general Multiple Demand (MD) network more than standard reading. This highlights the interplay between effort and language processing.
  • Time is of the essence: The development of this shorter localizer is a practical advancement for fMRI research, making it more accessible and efficient, especially for populations with limited scanner time or tolerance.

Key Results

  • Time Savings: The speeded localizer task requires only a single run of 3.5 minutes, compared to the standard version which typically involves two runs totaling approximately 12 minutes.
  • Enhanced Contrast: The speeded localizer showed a significantly larger "sentences > nonwords" effect (β = 0.723, p < 0.0001) compared to the standard localizer. This was driven by higher responses to sentences (β = 0.293, p = 0.017) and lower responses to non-words (β = -0.430, p < 0.0001) in the speeded version.
  • Similar Topography: Activation maps from the speeded and standard localizers showed high similarity, with Fisher-transformed spatial correlations averaging 0.859 between the two versions. Dice overlap coefficients also indicated substantial overlap.
  • Language Selectivity Maintained: The language fROIs identified by both localizers showed no significant response to a demanding non-linguistic spatial working memory task, confirming their selectivity for language.
  • MD Network Engagement: The speeded reading task led to greater engagement of the Multiple Demand (MD) network during sentence processing compared to the standard version (speeded > standard; β = 0.200, p < 0.0001).

Limitations and Future Work

  • Reading Proficiency: The speeded localizer requires a certain level of reading proficiency, potentially making it less suitable for populations with reading difficulties, such as young children or individuals with certain types of aphasia.
  • Task Order Confound: The study design administered the standard localizer before the speeded localizer. While experimental materials differed, the potential for task order effects (practice, strategy changes) cannot be entirely ruled out. The authors note that the observed increase in sentence response in the speeded version, despite being later in the session, suggests this confound might not be a major issue, but it remains a point for future investigation.
  • MD Network Localization: A practical consequence of the speeded localizer is that the "nonwords > sentences" contrast, which is used to define MD regions in the standard localizer, is no longer effective. Researchers will need to use a separate task (like the spatial working memory task used in this study) to define MD fROIs if needed.

Future work could explore the optimal presentation speed for different populations and investigate the precise nature of the MD network's engagement during speeded reading. The authors also suggest that this efficient localizer could be integrated into multimodal approaches or used in studies relating neural measures to behavior and genetics.

Summary updated

Abstract

Abstract The field of human cognitive neuroscience is increasingly acknowledging inter-individual differences in the precise locations of functional areas and the corresponding need for individual-level analyses in fMRI studies. One approach to identifying functional areas and networks within individual brains is based on robust and extensively validated ‘localizer’ paradigms—contrasts of conditions that aim to isolate some mental process of interest. Here, we present a new version of a localizer for the fronto-temporal language-selective network. This localizer is similar to a commonly-used localizer based on the reading of sentences and nonword sequences (Fedorenko et al., 2010) but uses speeded presentation (200ms per word/nonword). Based on a direct comparison between the standard version (450ms per word/nonword) and the speeded versions of the language localizer in 24 participants, we show that a single run of the speeded localizer (3.5 min) is highly effective at identifying the language-selective areas: indeed, it is more effective than the standard localizer given that it leads to an increased response to the critical (sentence) condition and a decreased response to the control (nonwords) condition. This localizer may therefore become the version of choice for identifying the language network in neurotypical adults or special populations (as long as they are proficient readers), especially when time is of essence.
BibTeX
@misc{tuckute2024minute,
  author = {Greta Tuckute and Elizabeth Jiachen Lee and Aalok Sathe and Evelina Fedorenko},
  title = {{A 3.5-minute-long reading-based fMRI localizer for the language network}},
  year = {2024},
  doi = {10.1101/2024.07.02.601683},
  howpublished = {bioRxiv},
}
Figure