Toward Pathology-guided Illumination Optimization for Neuromodulation with PhomiNeuro
- 1Institute of Medical Technology, Peking University Health Science Center, Beijing, China
Abstract
Clinical treatment planning for near-infrared (NIR) neuromodulation requires patient-specific dosimetry to optimize light fluence (LF) delivery to cortical targets. The gold-standard Monte Carlo photon-transport forward solver is accurate but computationally expensive and non-differentiable for personalized inverse design across subjects. Here, we present PhomiNeuro, a foundation-model-encoded differentiable surrogate for time-resolved LF modeling and pathology-guided inverse design. A pretrained 3D medical imaging foundation model (VISTA3D) was domain-adapted to 285 training head models annotated with optical properties, and then coupled to an implicit neural representation (INR) that predicts LF at arbitrary spatiotemporal coordinates. Regularized by a time-dependent diffusion-equation residual, it demonstrated superior fidelity in 84 held-out participants, and ablations identified VISTA3D-derived anatomical priors as dominant and physics regularization as complementary. PhomiNeuro enables fast and robust queries of LF and its gradients with respect to illumination parameters for targets guided by individual structural magnetic resonance imaging and amyloid positron emission tomography, achieving 466-fold per-iteration speedup and 167-fold end-to-end optimization speedup. Explanatory analyses of LF delivery variability, including sex differences and amyloid status, identified structural variations as the primary drivers. These results position PhomiNeuro as a highly extensible translational framework toward personalized treatment planning and digital twin development in precision neuromodulation.
Related
- Noninvasive ultrashort wave neuromodulation enhances glymphatic clearance to alleviate traumatic brain edema and Alzheimer’s pathology
- MR-Guided Focused Ultrasound Neuromodulation of Nucleus Accumbens to Attenuate Oral Fentanyl Motivation in A Self-Administration Rat Model
- Bridging the Modality Bottleneck in Pathology MIL through Molecularly Guided Feature Adaptation
- Active-learning-guided optimization of cell-free systems for genome-wide transcriptomic profiling reveals progressive layers of regulation
- MFG: Multi-turn Feedback Guided Molecular Optimization
BibTeX
@misc{dong2026toward,
title = {{Toward Pathology-guided Illumination Optimization for Neuromodulation with PhomiNeuro}},
author = {S. Dong and M. Guan and L. Yang and G. Liu and A. Rominger and Y. Yuan and W. Ren and R. Ni and X. Wei},
howpublished = {bioRxiv},
year = {2026},
month = jul,
doi = {10.64898/2026.07.04.736179},
url = {https://www.biorxiv.org/content/10.64898/2026.07.04.736179v2},
note = {CC BY-NC-ND},
}
RIS
TY - UNPB
TI - Toward Pathology-guided Illumination Optimization for Neuromodulation with PhomiNeuro
AU - S. Dong
AU - M. Guan
AU - L. Yang
AU - G. Liu
AU - A. Rominger
AU - Y. Yuan
AU - W. Ren
AU - R. Ni
AU - X. Wei
PY - 2026
DA - 2026/07/04
PB - bioRxiv
DO - 10.64898/2026.07.04.736179
UR - https://www.biorxiv.org/content/10.64898/2026.07.04.736179v2
AB - Clinical treatment planning for near-infrared (NIR) neuromodulation requires patient-specific dosimetry to optimize light fluence (LF) delivery to cortical targets. The gold-standard Monte Carlo photon-transport forward solver is accurate but computationally expensive and non-differentiable for personalized inverse design across subjects. Here, we present PhomiNeuro, a foundation-model-encoded differentiable surrogate for time-resolved LF modeling and pathology-guided inverse design. A pretrained 3D medical imaging foundation model (VISTA3D) was domain-adapted to 285 training head models annotated with optical properties, and then coupled to an implicit neural representation (INR) that predicts LF at arbitrary spatiotemporal coordinates. Regularized by a time-dependent diffusion-equation residual, it demonstrated superior fidelity in 84 held-out participants, and ablations identified VISTA3D-derived anatomical priors as dominant and physics regularization as complementary. PhomiNeuro enables fast and robust queries of LF and its gradients with respect to illumination parameters for targets guided by individual structural magnetic resonance imaging and amyloid positron emission tomography, achieving 466-fold per-iteration speedup and 167-fold end-to-end optimization speedup. Explanatory analyses of LF delivery variability, including sex differences and amyloid status, identified structural variations as the primary drivers. These results position PhomiNeuro as a highly extensible translational framework toward personalized treatment planning and digital twin development in precision neuromodulation.
ER -