Diffusion-driven lensless fiber endomicroscopic quantitative phase imaging towards digital pathology

TL;DR

A speckle-conditioned diffusion model (SpecDiffusion), which reconstructs phase images directly from speckles captured at the detection side of a multi-core fiber (MCF) and shows powerful generalization capacity for unseen objects, such as test charts and biological tissues.

Abstract

Lensless fiber endomicroscope is an emerging tool for in-vivo microscopic imaging, where quantitative phase imaging (QPI) can be utilized as a label-free method to enhance image contrast. However, existing single-shot phase reconstruction methods through lensless fiber endomicroscope typically perform well on simple images but struggle with complex microscopic structures. Here, we propose a speckle-conditioned diffusion model (SpecDiffusion), which reconstructs phase images directly from speckles captured at the detection side of a multi-core fiber (MCF). Unlike conventional neural networks, SpecDiffusion employs iterative phase denoising steps for speckle-driven phase reconstruction. The iteration scheme allows SpecDiffusion to break down the phase reconstruction process into multiple steps, gradually building up to the final phase image. This attribute alleviates the computation challenge at each step and enables the reconstruction of rich details in complex microscopic images. To validate its efficacy, we build an optical system to capture speckles from MCF and construct a dataset consisting of 100,000 paired images. SpecDiffusion provides high-fidelity phase reconstruction results and shows powerful generalization capacity for unseen objects, such as test charts and biological tissues, reducing the average mean absolute error of the reconstructed tissue images by 7 times. Furthermore, the reconstructed tissue images using SpecDiffusion shows higher accuracy in zero-shot cell segmentation tasks compared to the conventional method, demonstrating the potential for further cell morphology analysis through the learning-based lensless fiber endomicroscope. SpecDiffusion offers a precise and generalized method to phase reconstruction through scattering media, including MCFs, opening new perspective in lensless fiber endomicroscopic imaging.

Figures (6)

• Figure 1: Illustration demonstrates the working principle of the diffusion-driven lensless fiber endomicroscope. The quantitative phase reconstruction process involves using speckle images captured at the detection system to guide the denoising process of the SpecDiffusion model. The reconstructed phase images are applicable to digital pathology tasks, including cell segmentation, enhancing both accuracy and detail in diagnostic imaging.
• Figure 2: Architecture of the speckle-conditioned diffusion model (SpecDiffusion). a Training process of SpecDiffusion. Phase label is mixed with Gaussian noise, and SpecDiffusion is trained to predict the imposed noise with the guidance of speckle. b Inference process of SpecDiffusion. With the guidance from input speckle, the randomly-generated initial phase is gradually denoised and transformed towards label phase by SpecDiffusion.
• Figure 3: Diffusion-driven phase reconstruction of ImageNet images through lensless fiber endomicroscope. a Ground truth phase images. b Speckle patterns captured at the detection side of the lensless fiber endomicroscope. c and d Reconstructed phase images by U-Net and SpecDiffusion. The SSIM value for each reconstructed image with respect to its corresponding phase label is shown. Scale bars $50~ \mu m$.
• Figure 4: Diffusion-driven phase reconstruction of test chart through the lensless fiber endomicroscope. a Speckle pattern captured at the detection side of lensless fiber endomicroscope . b Ground truth images of the test chart. c U-Net reconstructed phase image. d SpecDiffusion reconstructed phase image. e Phase reconstruction contrast of ground truth, U-Net and SpecDiffusion. Scale bars $50~ \mu m$.
• Figure 5: Diffusion-driven cancer tissue reconstruction through the lensless fiber endomicroscope. a Ground truth phase images. b Speckle patterns from MCF captured at the detection side of the lensless fiber endomicroscope. c and d Reconstructed phase image by U-Net and SpecDiffusion. e MAE, f PSNR, g SSIM and h 2D correlation coefficient distribution evaluated on $1,000$ reconstructed tissue images by U-Net and SpecDiffusion. i MAE, j PSNR, k SSIM and l 2D correlation coefficient evaluated on $1,000$ reconstructed tissue images by U-Net and SpecDiffusion with varying tissue dataset size. The SSIM value for each reconstructed image with respect to its corresponding phase label is shown. Scale bars $50~ \mu m$.