NeuralRTE: Forward photon transport simulations in turbid media aided by machine learning and artificial intelligence

NeuralRTE introduces a pilot study of a learning-based algorithm for photon transport in turbid media, which is crucial in Biomedical Optics. By leveraging emerging artificial intelligence techniques, our approach produces accurate estimates of light propagation through scattering, tissue-like media. The developed solutions are open-source and power-efficient, making them suitable for embedded applications. We investigate the capabilities and limitations of machine learning for forward simulations and compare their accuracy and performance with well-known and established methodologies. A modular, comprehensive dataset generation, training, and testing pipeline is presented, and direct comparisons with ground truth analytical and computational results have been conducted. Our findings demonstrate the potential for high efficacy, speed, and accuracy of the proposed approaches. Moreover, we showcase practical examples of 3D digital twin rendering of melanoma-affected human skin using the developed techniques.