Spatial resolution of parameterized temporal data in 𝑙0-norm-based time-domain diffuse optical tomography (TD-DOT)

Due to the advancements of a time domain system [1-3], time-domain diffuse optical tomography (TD-DOT) has witnessed a revival in recent years. To improve reconstruction quality, various time data types have been explored, which include the full time of flight (ToF) data, gated time data, various moments from the Mellin transform, and its combination with the Laplace transform. It is well known that the full temporal data gives the highest accuracy to reconstruction, but in practice, it suffers from being sensitive to measurement noise. For this reason, parameterized temporal data, also called temporal filters, is often used as an alternative to the ToF data as demonstrated to be more robust to noise while providing reasonable accuracy with computational efficiency [4-5]. Recently, our group reported the l0-norm-based SENSOR algorithm [6] that achieves 1mm spatial resolution at a depth of 6mm within 20 milliseconds from using three parameterized temporal data types – zero-moment 𝐸, 1^st moment 𝑀¹ and exponential feature 𝐿(𝑠). In this work, we focus on evaluating the performance of the exponential feature 𝐿(𝑠) with respect to their spatial resolution as applied to absorption targets located at various depths beyond 10 mm within the tissue by using the 𝑙₀-norm minimization.