High-dynamic-range fluorescence confocal imaging with superconducting nanowire single-photon detectors and deep learning enhancement

We present a high-dynamic-range fluorescence imaging method using a superconducting nanowire single-photon detector (SNSPD) integrated with a custom confocal microscope. While SNSPDs offer exceptional quantum efficiency and low dark counts, their limited reset time constrains image dynamic range. To address this, we developed a convolutional neural network to enhance 8-bit SNSPD images. We trained the model on indocyanine green-stained cells images captured with low excitation power (~500 μW) and short point exposure (4 μs). The model successfully recovered shadow and highlight details, enabling HDR reconstruction. This approach significantly improves image contrast, reduces photobleaching, and demonstrates the potential of SNSPDs for rapid, low-light biomedical imaging.