Improving DOT reconstruction accuracy without MRI: Practical guidelines from systematic head model evaluations
Imaging Neuroscience · 12.11.2026 · 4
DOI: 10.1162/imag.a.1445 ↗Kurz gefasst
Leitlinie in der Zeitschrift Imaging Neuroscience, online erschienen am 12.11.2026. Verfasst von Nils Harmening und 2 weiteren Autor:innen. Beteiligte Einrichtung: BIFOLD - Berlin Institute for the Foundations of Learning and Data, Berlin, Germany. Hauptthemen laut MEDLINE: Head, Brain, Tomography, Optical, Image Processing, Computer-Assisted. Die Arbeit nennt 50 Literaturangaben. Der Volltext ist frei zugänglich (Open Access, CC BY 4.0). Gefördert von: Bundesministerium für Bildung und Forschung.
Abstract (Originalsprache)
High-density diffuse optical tomography (HD-DOT) is a promising modality for functional brain imaging, usable as a complement or an alternative to fMRI. Head models are commonly built from participant-specific MRI scans, but MRI access is often exactly what is missing in the settings where HD-DOT is most useful. We, therefore, systematically evaluated localization accuracy across a range of head models—subject-specific MRI-based, photogrammetry-based individualized, and atlas-based generic models—using simulated diffuse optical tomography (DOT) data from 15 participants. Our results demonstrate that individualized head models derived from photogrammetry scans lower the median peak localization error by about 3.7 mm relative to the atlas models at every tested probe density, and narrow the reconstructed extent by 7.6 mm at the sparse probe down to 3.1 mm at high density. Across the four tested levels, the spatially variant regularization parameter moves the median peak error by up to 7 mm, comparable with the difference between the best and the worst head model. Its optimum is more stable as a rank in the probe’s sensitivity distribution than as a fixed sensitivity value. We further analyze how localization accuracy relates to probe density and cortical region. By offering a comprehensive resource that maps reconstruction performance onto standard MNI coordinates and cortical parcels, this work delivers practical guidelines for selecting optimal head modeling strategies in HD-DOT studies when subject-specific MRIs are not available.
Schlagwörter
MeSH-Schlagwörter (MEDLINE)
Stichwörter der Autor:innen
Autor:innen und Institutionen
- BIFOLD - Berlin Institute for the Foundations of Learning and Data, Berlin, Germany.
- Intelligent Biomedical Sensing (IBS) Lab, Machine Learning Department, Technical University of Berlin, Berlin, Germany.
- Boston University, Neurophotonics Center, Department of Biomedical Engineering, Boston, MA, United States.
ORCID-Profile: Nils Harmening, David A Boas, Alexander von Lühmann
Förderung
- Bundesministerium für Bildung und Forschung
Zitieren
APA
Harmening, N., Boas, D. A., & von Lühmann, A. (2026). Improving DOT reconstruction accuracy without MRI: Practical guidelines from systematic head model evaluations. Imaging Neuroscience, 4. https://doi.org/10.1162/imag.a.1445
Vancouver
Harmening N, Boas DA, von Lühmann A. Improving DOT reconstruction accuracy without MRI: Practical guidelines from systematic head model evaluations. Imaging Neuroscience. 2026;4. doi:10.1162/imag.a.1445
BibTeX
@article{Harmening2026,
author = {Harmening, Nils and Boas, David A and von Lühmann, Alexander},
title = {Improving DOT reconstruction accuracy without MRI: Practical guidelines from systematic head model evaluations},
journal = {Imaging Neuroscience},
year = {2026},
volume = {4},
doi = {10.1162/imag.a.1445}
}
EUVIMED