Enhancing low-light endoscopic laser speckle contrast imaging using the noise-correction framework of an analog gain
None:
Endoscopic laser speckle contrast imaging (eLSCI) has the potential to reduce the risk of anastomotic leakage in minimally invasive colorectal cancer resection surgeries. However, its clinical utility is limited by quantization distortion, low SNR, low sensitivity, and flow imaging bias-all mainly due to low light intensity, detector noise, and uneven laser illumination. While analog gain amplification can enhance quantization accuracy and speckle dynamic range, the inherent trade-off between signal enhancement and noise amplification renders analog gain less favorable for LSCI. Moreover, the imaging community lacks a consensus on the noise model under analog gain. To address these challenges, we established a noise-correction framework under analog gain based on synthesis noise and maximum likelihood estimation (AGNc-SNMLE). AGNc-SNMLE balances the enhancement of signal intensity with the correction of amplified noise introduced by analog gain. Simulation and experiments showed that AGNc-SNMLE significantly improved SNR and sensitivity, optimized the imaging signal-to-background ratio, expanded flow linearity, and reduced flow imaging bias of eLSCI under low-light intensity. AGNc-SNMLE requires no hardware modifications or additional computational overhead, and is fully compatible with low-cost laser and camera, making it a powerful and cost-effective solution for low-light intensity eLSCI.
