Machine learning and the Gegenbauer Kernel improve mapping of sub-diffuse optical properties in the spatial frequency domain

Imaging sub-diffuse optical properties of cancerous and normal skin tissue using machine learning-aided spatial frequency domain imaging

“… 32 In order to allow for simulation across a wide range of the phase function parameters and , our look-up table is generated with the modified Henyey-Greenstein 23 (MHG) function when and fall within MHG’s valid range ( 31 ) and is otherwise generated with the GK. 51 We approximated GK’s valid range to be and 38 (Supplement 1 in the Supplementary Material ), so our method cannot sample phase functions with . This is not a concerning limitation given the range of phase function parameters seen in our phantoms and in biological tissue.…”

“…This is not a concerning limitation given the range of phase function parameters seen in our phantoms and in biological tissue. 2 , 21 , 26 , 32 – 35 Our method uses a novel mapping which relates the and values to the GK parameters 38 (Supplement 1 in the Supplementary Material ) to calculate the GK parameters when using the GK phase function. It should be noted that the GK phase function has an analytical inverse CDF 52 that can be used in tandem with our mapping in place of the inverse CDF look-up table, but using look-up tables for both phase functions allowed for more seamless switching between GK and MHG, the latter of which does not have an analytical inverse CDF.…”

“… 36 , 37 However, while this method was shown to be successful for simulated data, this method has never been calibrated for and applied to physical experimental data. Moreover, Naglič et al.’s model was trained on data simulated using the Gegenbauer kernel (GK) phase function approximation, whose domain 38 also does not cover the full range of optical properties seen in biological tissue. 2 , 21 , 26 , 32 – 35 …”