Improved Diffuse Optical Tomography (DOT) via Multi-Layer Perceptron (MLP) Using Single-Wavelength Data

Diffuse Optical Tomography (DOT) is a non-invasive imaging technique that uses near-infrared (NIR) light to image tissues. This method provides information about the tissue’s optical properties such as absorption and scattering which relate to its physiological characteristics (e.g., oxygenation, hemoglobin content). DOT suffers from low spatial resolution because light in tissue undergoes multiple scattering. Hence, high-resolution reconstruction of images from DOT data remains a challenging task. In multi-wavelength DOT, light of different wavelengths is used to obtain more detailed information about the tissue composition as different biological chromophores (e.g., oxygenated and deoxygenated hemoglobin, water, lipids) absorb light differently at different wavelengths. In this work, we propose a novel approach using a Multi-Layer Perceptron (MLP) neural network to improve the resolution and accuracy of DOT reconstructions by leveraging single-wavelength data. Our method demonstrates improvement in resolving capabilities and the quality of reconstructed images making it suitable for medical applications including breast cancer detection and brain imaging. The multi-wavelength data improves the characterization of tissue properties and enhances the separation of chromophore concentrations leading to more accurate reconstructions.