TL;DR
This paper introduces the "New York Head" (ICBM-NY), a highly detailed finite element model (FEM) of the human head that significantly improves the accuracy of EEG source localization and transcranial electrical stimulation (tES) targeting compared to existing standard models. This new model aims to be a widely adopted standard for researchers when individual MRI data is unavailable.
What Problem Does This Paper Address?
Accurate modeling of how electrical currents flow through the human head is crucial for both electroencephalography (EEG) for "reading" brain activity and transcranial electrical stimulation (tES) for "writing" to the brain. Existing models, primarily boundary element models (BEMs), are computationally efficient but lack anatomical detail, especially regarding the cerebrospinal fluid (CSF) and the full extent of the head and neck. Finite element models (FEMs) offer higher precision but are computationally expensive and require detailed individual anatomical segmentation, which is often impractical. Current standard models are often based on an "arbitrary" individual (like the Colin27 head) or a less detailed average (like ICBM152), leading to inaccuracies when applied to diverse populations or when precise targeting is needed. This paper addresses the need for a highly accurate, yet standardized, volume conductor model that balances anatomical detail with practical usability.
Methods
The authors developed the "New York Head" (ICBM-NY) by combining detailed anatomical data from multiple sources. They used the high-resolution brain data from the ICBM152 v2009b template and the non-brain region data from the ICBM152 v6 template. Crucially, they extended the field of view to include the neck using an additional average head dataset. This composite model was then segmented into six tissue types (scalp, skull, CSF, gray matter, white matter, air cavities) at a high resolution of 0.5 mm³. A finite element model (FEM) was generated from this detailed segmentation. To facilitate widespread use, the lead fields (which describe the relationship between current sources and scalp potentials) were precomputed for 231 electrode locations and made available. For evaluation, they compared the ICBM-NY model against several other models, including FEMs of individual subjects (considered "ground truth"), BEMs of the ICBM152 template, individualized BEMs, and FEMs of other individuals. Performance was assessed using metrics related to lead field approximation accuracy, EEG source localization error, and tES targeting accuracy.
Datasets
The primary dataset used for creating the ICBM-NY model was derived from the ICBM152 anatomical template (specifically, the ICBM152 v2009b for brain and ICBM152 v6 for non-brain regions). The field of view was extended to the neck using an additional average head dataset of 26 subjects. For evaluation, the authors used high-resolution MRI scans from four individual subjects (INDV1-4) as the "ground truth" reference FEMs.
Key Insights and Ideas
- The CSF layer is critical: The inclusion of the cerebrospinal fluid (CSF) layer, often omitted in simpler BEM models, is shown to be important for accurate current flow modeling, particularly for currents directed perpendicular to the cortical surface.
- Extended Field of View Matters: Including the neck region in the model significantly impacts current flow patterns, especially for tES targeting where distant reference electrodes are common.
- FEMs offer superior detail: While BEMs are computationally efficient, FEMs, when built with sufficient anatomical detail, provide a more accurate representation of current flow, leading to better EEG source localization and tES targeting.
- A standardized, detailed FEM is achievable: The ICBM-NY demonstrates that it's possible to create a highly detailed, yet standardized, FEM that balances anatomical realism with practical usability, making it a valuable tool when individual MRIs are not available.
Key Results
The ICBM-NY model demonstrated superior performance compared to standard BEM models of the ICBM152 anatomy and FEMs of mismatched individual anatomies. Specifically:
- EEG Source Localization: The ICBM-NY achieved a median localization error of 10.3 mm, outperforming INCG FEM (13.3 mm) and ICBM BEM (10.8 mm).
- tES Targeting Accuracy: The ICBM-NY outperformed INCG FEM, ICBM BEM, and WARP BEM in terms of both relative error in electric field intensity and spatial mismatch of peak intensity.
- Lead Field Approximation: The ICBM-NY showed competitive performance in lead field approximation, outperforming incongruent FEMs and BEMs, though it was slightly outperformed by BEMs that utilized the reference anatomy directly.
Limitations and Future Work
- Demographic Generalizability: The current evaluation is based on four Caucasian male subjects. Further studies with larger and more diverse reference heads are needed to confirm the ICBM-NY's generalizability to the broader population.
- WM Anisotropy and Skull Detail: The ICBM-NY model does not include white matter anisotropy or differentiate between skull spongiosa and compacta due to a lack of DTI data for the ICBM152 template. While these factors can influence current flow, their necessity is still debated, and incorporating them would add significant complexity. The authors suggest that future work could incorporate anisotropy by registering DTI data from an individual, though this might introduce noise.
- Electrode Modeling: For tES, electrodes were modeled as point-like sources rather than realistic pads. While this offers flexibility for various montages, it might not perfectly capture the behavior of specific electrode types.
- Future Work: The authors propose the ICBM-NY as a new standard head model for tES targeting and EEG source localization when individual MRI data is unavailable. They have made all model data available online to facilitate broad adoption and future research. Future work could focus on incorporating more detailed anatomical features like white matter anisotropy and validating the model with experimental data.
Summary updated