Presurgical resting‐state functional MRI language mapping with seed selection guided by regional homogeneity

Abstract

Purpose Resting‐state functional MRI (rs‐FMRI) has shown potential for presurgical mapping of eloquent cortex when a patient’s performance on task‐based FMRI is compromised. The seed‐based analysis is a practical approach for detecting rs‐FMRI functional networks; however, seed localization remains challenging for presurgical language mapping. Therefore, we proposed a data‐driven approach to guide seed localization for presurgical rs‐FMRI language mapping. Methods Twenty‐six patients with brain tumors located in left perisylvian regions had undergone task‐based FMRI and rs‐FMRI before tumor resection. For the seed‐based rs‐FMRI language mapping, a seeding approach that integrates regional homogeneity and meta‐analysis maps (RH+MA) was proposed to guide the seed localization. Canonical and task‐based seeding approaches were used for comparison. The performance of the 3 seeding approaches was evaluated by calculating the Dice coefficients between each rs‐FMRI language mapping result and the result from task‐based FMRI. Results With the RH+MA approach, selecting among the top 6 seed candidates resulted in the highest Dice coefficient for 81% of patients (21 of 26) and the top 9 seed candidates for 92% of patients (24 of 26). The RH+MA approach yielded rs‐FMRI language mapping results that were in greater agreement with the results of task‐based FMRI, with significantly higher Dice coefficients (P < .05) than that of canonical and task‐based approaches within putative language regions. Conclusion The proposed RH+MA approach outperformed the canonical and task‐based seed localization for rs‐FMRI language mapping. The results suggest that RH+MA is a robust and feasible method for seed‐based functional connectivity mapping in clinical practice.

The role of resting-state functional MRI for clinical preoperative language mapping

“…In a recently published research study performed at our institution, we compared SBC rs-fMRI maps generated using tb-fMRI activations (apart from tumor) as seed points, an automated ReHo method, and a canonical (anatomical landmark) approach. The ReHo method yielded rs-fMRI language mapping results that were in greater agreement with the results of tb-fMRI, with significant higher Dice coefficients (p < 0.5) than that of tb-fMRI and canonical approaches within the putative language areas [19]. Based on the results of this study, we plan to formally integrate an automated ReHo method to guide SBC rs-fMRI seed placement into our clinical workflow in the near future.…”

Identification of the Language Network from Resting-State fMRI in Patients with Brain Tumors: How Accurate Are Experts?

“…Rs-fMRI is being increasingly used in the setting of presurgical brain mapping. 3,14,15 Despite the potential subject level variability of data accuracy, 6,7 nevertheless in select cases, rs-fMRI may be considered a viable option for presurgical brain mapping; indeed, at least 1 institution has included rs-fMRI in their presurgical brain mapping paradigm without obtaining task-fMRI. 3,16 For this purpose, obtaining highly accurate intrinsic brain network data is paramount to avoid adverse outcomes following surgery.…”

Comparative analysis of brain language templates with primary language areas detected from presurgical fMRI of brain tumor patients

“…Four language templates that had previously shown success for applications in presurgical language mapping were selected for this study: (1) anatomically determined with Harvard–Oxford Atlas (Caviness et al., 1996 ) (Template A), (2) generated based on tb‐fMRI study (Branco et al., 2016 ; Fedorenko et al., 2010 ) (Template B), (3) generated based on rs‐fMRI study (Shirer et al., 2012 ; Zacà et al., 2018 ) (Template C), and (4) obtained from meta‐analysis results from Neurosynth with anatomical constraints (Hsu et al., 2020 ; Yarkoni et al., 2011 ) (Template D) (Figure 1 ). More details about the construction of these atlas can be found in Section 4 .…”

“…Language brain templates typically include anterior and posterior primary language areas (PLAs), and some include other language‐associated areas that are representative of a population of subjects (Fedorenko et al., 2010 ). After spatially transformed to an individual subject's space, such templates can help clinical fMRI analysis to evaluate language laterality from tb‐ or rs‐fMRI (Agarwal et al., 2018 ; Gohel et al., 2019 ; Pillai & Zaca, 2011 ; Ruff et al., 2008 ) and to guide seed selection for mapping the language network with rs‐fMRI (Hsu et al., 2020 ). In addition, previous studies have demonstrated the feasibility of categorizing functional networks from independent component analysis (ICA) of rs‐fMRI by using an automated template‐matching process and showed success in identifying language networks for presurgical mapping (Branco et al., 2016 ; Tie et al., 2014 ).…”

“…These causes of anatomic variability in the language function highlight the difficulty of finding a template for presurgical fMRI applications. To address this difficulty, this study aimed to evaluate and compare four publicly available language templates that have been used in previously published clinical fMRI studies (Branco et al., 2016 ; Gohel et al., 2019 ; Hsu et al., 2020 ; Zacà et al., 2018 ). Each template was analyzed for its ability to capture the maximum amount of PLA activation from presurgical fMRI of brain tumor patients, while minimizing the inclusion of activations that are not in the PLAs.…”