DeepPD: Joint Phase and Object Estimation from Phase Diversity with Neural Calibration of a Deformable Mirror

2025-04-19

AI Summary

Fluorescence microscopy is limited by wavefront aberrations introduced by the specimen and optical system, degrading spatial resolution. Conventional phase diversity (PD) methods suffer from reliance on Zernike polynomial parameterization, requirements for numerous image acquisitions, or precise calibration of deformable mirrors (DMs). This paper proposes DeepPD: the first framework embedding a neural DM calibration model into a deep learning architecture to jointly reconstruct the object—represented via implicit neural representation—and the wavefront—parameterized using a NeRF-inspired phase field—from only five PD images. DeepPD eliminates Zernike basis constraints and integrates a physics-driven forward model of light propagation, ensuring robustness under severe aberrations. Experiments demonstrate diffraction-limited resolution recovery on both calibration targets and PtK2 cell samples stained for myosin immunofluorescence, with reconstruction fidelity significantly surpassing conventional PD approaches.

Technology Category

• Machine Learning: Calibration & Uncertainty Quantification
• Computer Vision: Learning & Optimization for CV
• Search and Optimization: Learning to Search

Application Category

• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• User Modeling, Personalization and Recommendation: On-Device user modeling, personalization, and recommendation
• Graph Algorithms and Modeling for the Web: Graph neural networks and deep learning approaches for Web-related graphs

Abstract

Sample-induced aberrations and optical imperfections limit the resolution of fluorescence microscopy. Phase diversity is a powerful technique that leverages complementary phase information in sequentially acquired images with deliberately introduced aberrations--the phase diversities--to enable phase and object reconstruction and restore diffraction-limited resolution. These phase diversities are typically introduced into the optical path via a deformable mirror. Existing phase-diversity-based methods are limited to Zernike modes, require large numbers of diversity images, or depend on accurate mirror calibration--which are all suboptimal. We present DeepPD, a deep learning-based framework that combines neural representations of the object and wavefront with a learned model of the deformable mirror to jointly estimate both object and phase from only five images. DeepPD improves robustness and reconstruction quality over previous approaches, even under severe aberrations. We demonstrate its performance on calibration targets and biological samples, including immunolabeled myosin in fixed PtK2 cells.

Problem

• Overcoming sample-induced aberrations in fluorescence microscopy
• Jointly estimating object and phase from few images
• Improving robustness and quality in phase diversity methods

Innovation

• Neural representations for object and wavefront
• Learned model of deformable mirror calibration
• Joint phase and object estimation from five images