A systematic review of neuroimaging approaches to single-subject localization of language in the brain
Background. Task-based functional MRI has become the method of choice for researchers studying functional localization in the human brain. However, for a deeper understanding of brain function beyond group level generalizations, it is crucial to account for the reliability of brain activations in a single subject. Individual differences can influence group results in a multitude of ways and consequently lead to the mischaracterization of functional organization. Such errors can be detrimental to the accuracy of both basic research and clinical prognosis. Methodology. We performed a systematic review with the goal of understanding the state of the literature pertaining to mapping language regions using fMRI in individual participants. A thorough database search was carried out on published literature through April 2020. Results. Out of 977 papers identified through our literature search, 121 met our inclusion criteria for reporting single-subject fMRI results. Of these, 20 papers reported using single-subject level test-retest as a reliability measure. Among these papers, overlap measures such as Dice coefficient, Intraclass Correlation Coefficient, Euclidean Distance between peak activation or center of mass, or Receiver Operating Characteristic were used to further quantify the variability in their results. Among other categories, papers focused on comparing performance between language tasks, multimodal validation of fMRI results, technical development of protocols and clinical case studies on specific disease conditions. Conclusion. Incorporating reliability and validity measures in language mapping paradigms increases the likelihood that task-based activations in the brain are reproducible. However, very few papers reported measures of test-retest reliability. In the absence of quantified reproducibility, results from paradigms used for single-subject language mapping may need to be treated with caution. Future attempts to optimize the localization of language networks in individuals will benefit from the broader adoption of reliability metrics for different tasks and acquisition parameters.