Deep-learning-based separation of shallow and deep layer blood flow rates in diffuse correlation spectroscopy

Area of Science:

• Biomedical Optics
• Physiological Measurement
• Machine Learning in Healthcare

Background:

• Diffuse correlation spectroscopy (DCS) struggles to differentiate cutaneous and deep tissue blood flow.
• Accurate quantification of blood flow in different tissue layers is crucial for physiological monitoring.

Purpose of the Study:

• To develop a method for directly quantifying shallow and deep-layer tissue blood flow rates.
• To overcome the limitations of conventional diffuse correlation spectroscopy in distinguishing blood flow contributions.

Main Methods:

• Utilized a long short-term memory (LSTM) network, a type of recurrent neural network.
• Exploited differential contributions of shallow and deep-layer flow rates to auto-correlation functions.
• Validated the model using a two-layer flow phantom experiment.

Main Results:

• The LSTM network accurately predicted shallow and deep-layer flow rates.
• Achieved high accuracy with RMSE values of 0.047 and 0.034 ml/min/100g, and R² values of 0.99 for both layers.
• Demonstrated successful differentiation of blood flow in simulated two-layer tissue.

Conclusions:

• The proposed LSTM network offers a robust solution for separating and quantifying cutaneous and deep tissue blood flow.
• This technique is valuable for assessing blood flow dynamics in active muscles during physiological changes like exercise.
• Enables more precise evaluation of localized blood flow responses in complex biological tissues.