Illumination coding meets uncertainty learning: toward reliable AI-augmented phase imaging
MetadataShow full item record
Citation (published version)Yujia Xue, Shiyi Cheng, Yunzhe Li, Lei Tian. "Illumination coding meets uncertainty learning: toward reliable AI-augmented phase imaging." arXiv,
We propose a physics-assisted deep learning (DL) framework for large space-bandwidth product (SBP) phase imaging. We design an asymmetric coded illumination scheme to encode high-resolution phase information across a wide field-of-view. We then develop a matching DL algorithm to provide large-SBP phase estimation. We show that this illumination coding scheme is highly scalable in achieving flexible resolution, and robust to experimental variations. We demonstrate this technique on both static and dynamic biological samples, and show that it can reliably achieve 5X resolution enhancement across 4X FOVs using only five multiplexed measurements -- more than 10X data reduction over the state-of-the-art. Typical DL algorithms tend to provide over-confident predictions, whose errors are only discovered in hindsight. We develop an uncertainty learning framework to overcome this limitation and provide predictive assessment to the reliability of the DL prediction. We show that the predicted uncertainty maps can be used as a surrogate to the true error. We validate the robustness of our technique by analyzing the model uncertainty. We quantify the effect of noise, model errors, incomplete training data, and "out-of-distribution" testing data by assessing the data uncertainty. We further demonstrate that the predicted credibility maps allow identifying spatially and temporally rare biological events. Our technique enables scalable AI-augmented large-SBP phase imaging with dependable predictions.