Comparison of diffuse correlation spectroscopy analytical models for cerebral blood flow measurements
Abstract
Multi-layer diffuse correlation spectroscopy (DCS) models have been developed to reduce the contamination of superficial signals in cerebral blood flow index (CBFi) measurements. However, a systematic comparison of these models and clear guidance on model selection are still lacking. This study compares three DCS analytical models: semi-infinite, two-layer, and three-layer, focusing on their fitting strategies, performance, and suitability for CBFi and relative CBFi (rCBFi) estimation. We simulated DCS data using a four-layer slab head model with the Monte Carlo eXtreme (MCX) toolkit. Multiple fitting strategies were evaluated: early time lag range (ETLR) fitting with fixed or variable beta for the semi-infinite model, and single-distance (SD) and multi-distance (MD) fitting for the two- and three-layer models. Model performance was assessed based on CBFi sensitivity, accuracy of CBFi and rCBFi recovery, resistance to signal contamination from scalp and skull, sensitivity to assumed parameter errors, and computational efficiency across source-detector separations of 20 to 35 mm. Optimal fitting methods include ETLR with fixed beta for the semi-infinite model, SD with fixed beta for the two-layer model, and MD for the three-layer model. The multi-layer models achieved higher CBFi sensitivity (up to 100%) compared to 36.8% for the semi-infinite model. The two-layer model offered the best balance of accuracy and robustness, while the three-layer model enabled simultaneous recovery of CBFi, scalp BFi, and rCBFi. The semi-infinite model was the most computationally efficient, requiring only 0.38 seconds for 500 samples, supporting its use in real-time monitoring. This work offers a practical and systematic evaluation of DCS analytical models and provides guidance for selecting the most appropriate model based on application needs.