TY - GEN
T1 - Scalp Patch Based Diffuse Optical Tomography for Tumor Monitoring
AU - Mehra, Ayush
AU - Manavalan, Gokul
AU - Kupferman, Judy
AU - Arnon, Shlomi
N1 - Publisher Copyright:
2026 Published by SPIE.
PY - 2026/3/5
Y1 - 2026/3/5
N2 - Non-invasive brain monitoring without bulky instrumentation, ionizing radiation, or high cost remains a major challenge in neuroimaging. Diffuse Optical Tomography (DOT) offers a safe and portable solution by using near-infrared light to probe tissue absorption and scattering. However, conventional DOT systems rely on large, fiber-dense arrays that limit their suitability for continuous or bedside use. In this work, we present a compact, scalp-mounted optical patch with integrated light sources and detectors for localized brain interrogation. Photon propagation through multilayer head tissues is modeled using Monte Carlo (MCX) simulations to evaluate sensitivity to tumor-induced optical variations. A deep-learning classifier is trained on simulated fluence data to detect tumor presence. Four source–detector configurations (6 × 5, 4 × 3, 3 × 2, and 2 × 1) are investigated, yielding classification accuracies of 95.8 %, 94.5 %, 90.7 %, and 77.7 %, respectively. The results demonstrate the feasibility of a thin, wearable DOT patch for continuous, non-invasive brain tumor monitoring.
AB - Non-invasive brain monitoring without bulky instrumentation, ionizing radiation, or high cost remains a major challenge in neuroimaging. Diffuse Optical Tomography (DOT) offers a safe and portable solution by using near-infrared light to probe tissue absorption and scattering. However, conventional DOT systems rely on large, fiber-dense arrays that limit their suitability for continuous or bedside use. In this work, we present a compact, scalp-mounted optical patch with integrated light sources and detectors for localized brain interrogation. Photon propagation through multilayer head tissues is modeled using Monte Carlo (MCX) simulations to evaluate sensitivity to tumor-induced optical variations. A deep-learning classifier is trained on simulated fluence data to detect tumor presence. Four source–detector configurations (6 × 5, 4 × 3, 3 × 2, and 2 × 1) are investigated, yielding classification accuracies of 95.8 %, 94.5 %, 90.7 %, and 77.7 %, respectively. The results demonstrate the feasibility of a thin, wearable DOT patch for continuous, non-invasive brain tumor monitoring.
KW - Diffuse optical tomography
KW - Monte Carlo simulation
KW - brain imaging
KW - deep learning
UR - https://www.scopus.com/pages/publications/105035337967
U2 - 10.1117/12.3095245
DO - 10.1117/12.3095245
M3 - Conference contribution
AN - SCOPUS:105035337967
T3 - Progress in Biomedical Optics and Imaging - Proceedings of SPIE
BT - Neural Imaging and Sensing 2026
A2 - Fu, Ling
A2 - Ding, Jun
PB - SPIE
T2 - Neural Imaging and Sensing 2026
Y2 - 17 January 2026 through 18 January 2026
ER -