MCMLNet: Multilayer Monte Carlo Tissue-Optics Simulation and Surrogate Model Checkpoint Dataset

This dataset accompanies the MCMLNet repository, which provides neural surrogate models for accelerating multilayer Monte Carlo (MC) simulations of tissue reflectance. It includes large-scale simulated spectral reflectance data generated with flexible multilayer tissue models grounded in established tissue-optics literature, along with all corresponding surrogate-model checkpoints. The simulations were used to train, validate, and benchmark surrogate models that achieve MC-level accuracy while enabling inference up to five orders of magnitude faster. The dataset serves as a reusable community resource, reducing the need for repeated costly MC computations and supporting future work in spectral imaging, inverse problems, and AI-driven diagnostics. The dataset includes: Monte Carlo simulation outputs used to train the main surrogate model (>50M simulations across broad optical parameter ranges), as well as supplementary ablation datasets. Surrogate inference datasets, as used in the publication. Surrogate model checkpoints for our models, trained on different data qualities and scales, as well as for the reimplemented prior work surrogate models. Metadata describing optical device properties and chromophores. The corresponding codebase for training and inference is openly available at: https://github.com/IMSY-DKFZ/mcmlnet