Machine learning in multiexposure laser speckle contrast imaging can replace conventional laser Doppler flowmetry.

Area of Science:

• Biomedical Optics
• Physiological Measurement
• Medical Imaging

Background:

• Laser speckle contrast imaging (LSCI) offers high-speed blood flow visualization but suffers from nonlinear perfusion estimates dependent on exposure time.
• Laser Doppler flowmetry (LDF) provides a more reliable perfusion measure but operates at slower speeds.
• Bridging the speed and accuracy gap between LSCI and LDF is crucial for advanced hemodynamic monitoring.

Purpose of the Study:

• To develop and validate a novel method for accurate blood perfusion estimation using multi-exposure LSCI (MELSCI).
• To leverage artificial neural networks (ANN) for rapid and precise processing of MELSCI data into a perfusion metric comparable to LDF.
• To investigate the impact of noise on the accuracy of MELSCI-derived perfusion estimates.

Main Methods:

• Utilized seven exposure times (1-64 ms) in MELSCI to capture a range of speckle contrasts.
• Employed artificial neural networks (ANN) trained on simulated Doppler histograms and speckle contrasts from tissue models.
• Incorporated noise modeling into the ANN training process to enhance robustness.

Main Results:

• The ANN-based MELSCI method accurately estimated blood perfusion, achieving high correlation coefficients (R=1.000 for noise-free data, R=0.993 with noise, R=0.995 for in vivo data).
• Demonstrated the critical importance of accounting for noise in achieving reliable perfusion estimates.
• Successfully translated MELSCI data into LDF-equivalent perfusion values with high fidelity.

Conclusions:

• Artificial neural networks enable accurate conversion of MELSCI data to LDF-like perfusion estimates, overcoming LSCI's inherent nonlinearities.
• The developed MELSCI-ANN method offers a promising approach for high-speed, accurate blood perfusion monitoring in various biomedical applications.
• This technique enhances the utility of LSCI for quantitative physiological measurements.