Fig. 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.

Fig. 2. Architecture of the speckle-conditioned diffusion model (SpecDiffusion). (a) Training process of SpecDiffusion. The 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 toward the label phase by SpecDiffusion.

Fig. 3. Diffusion-driven phase reconstruction of ImageNet images through the 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 µm.

Fig. 4. Diffusion-driven phase reconstruction of the test chart through the lensless fiber endomicroscope. (a) Speckle pattern captured at the detection side of the 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 µm.

Fig. 5. Diffusion-driven cancer tissue reconstruction through the lensless fiber endomicroscope. (a) Ground truth phase images. (b) Speckle patterns from MCFs 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 distributions 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 µm.

Fig. 6. Cell segmentation by SAM on ground truth and reconstructed images. (a) Cell segmentation based on lensless MCF phase imaging by SAM. (b) Ground truth phase images and their cell segmentation results. (c) and (d) Reconstructed phase images by U-Net and SpecDiffusion, and their cell segmentation results. Scale bars: 50 µm.