Abstract

Title: MALT-NSOM: A Multi-Physics 3D Tensor Dataset for AI-Driven Oncological Biosensing and Targeted Therapeutics

Abstract:

This dataset provides a high-fidelity, multi-physics ground-truth benchmark for the MALT-NSOM (Metal-Assisted Lossless-Dielectric Nanophotonic Sub-diffraction Optical Microscopy) platform. Unlike noisy clinical datasets, this archive contains 3D tensor fields derived from first-principles Maxwell, Poisson, and Acoustic wave equations. It is designed for researchers developing machine learning models for predictive biomarker detection, sub-cellular diagnostic imaging, and non-invasive therapeutic targeting.

Data Generation & Methodology:

The dataset was constructed using three primary high-performance physics engines:

  1. Optics (MEEP/FDTD): 3D Maxwell tensor simulations of a bi-spherical Hafnium Oxide (HfO2) dimer cavity (Gap = 16.57 nm, Radius = 50.0 nm). The data includes 3D electromagnetic field distributions under 180 nm Deep-UV excitation, yielding a peak intensity enhancement factor of 20.69.

  2. Thermodynamics & PEMF Induction (FEniCS): Finite Element Analysis (FEA) of Poisson’s heat and magnetic induction equations. Includes datasets for temperature dissipation (Thermal delta +0.1478 Celsius) and induced voltage potentials from 15 Hz PEMF magnetic coils within a heterogeneous bone/marrow matrix.

  3. Acoustic Intervention (k-Wave): 3D acoustic tensor fields generated by a 24-element helical-conical array (2 MHz / 100 kHz multiplexing). Includes projection heatmaps of non-linear pressure gradients (peak 1.2 MPa) capable of transient Blood-Brain Barrier (BBB) modulation and non-thermal Glioblastoma (GBM) disruption.

Key Dataset Metrics:

  • Spatial Resolution: 1.0 - 2.0 pixels/nm (Optics), 0.2 mm (Acoustics).

  • Targeted Phenomena: Single-molecule (2.5 nm - 10 nm) biomarker refractive index perturbations.

  • Signal-to-Noise: +27.20 percent signal spike verification for label-free detection.

  • Synergy: Verified mechanical shearing energy density (2.11 normalized units) vs. thermal necrosis threshold (2.50).

Significance for AI Research:

This dataset provides "physics-perfect" training data for supervised and semi-supervised learning models. It eliminates the data-scarcity problem in label-free biosensing by providing explicit, noise-free ground truth for protein-photon interactions, heterogeneous acoustic wave refraction, and non-invasive therapeutic targeting. It serves as an ideal benchmark for models aiming to achieve sub-diffraction clinical diagnostics and automated in silico drug discovery.

Keywords: Nanophotonics, FDTD, Biophysics, Oncology, Glioblastoma, Biosensing, Computational Biology, Therapeutics Data Commons, Machine Learning, Ground Truth.

Creative Commons License

Creative Commons License
This work is licensed under a Creative Commons Attribution 4.0 License.

Download

Share

COinS