An artificial neural network model for accurate and efficient optical property mapping from spatial-frequency domain images

Conventional optical property mapping by spatial-frequency domain imaging (SFDI) is prone to relatively low efficiency due the iterative nature of nonlinear curve-fitting based on light transfer model, such as diffusion approximation equation. This study aims at expediating prediction with high accuracy developing an artificial neural network (ANN) model coupled SFDI. A dataset was first generated using Monte Carlo simulations Graphic Processing Unit. The ANN training then conducted and optimized for multiple factors (learning rate, batch size number neurons). Experiments simulation samples a solid phantom were carried out verify performance predicting mapping. Results showed that normalized mean absolute errors absorption coefficient (?a) reduced scattering (?'s) 0.18 % 0.027 %, while root square 0.01 0.14. Optical properties demonstrated proposed retained accuracy, about three orders magnitude faster than inverse model. Finally, tested measurement ‘Golden Delicious’ apples different bruising levels (non-, severe bruising). indicated could measure apple ?a ?'s accurately efficiently, measured capable detecting early in apples. Contrast bruise feature value between bruised non-bruised tissues significantly improved, enhancing detection.