[Paper Review] High-throughput fast full-color digital pathology based on Fourier ptychographic microscopy via color transfer

Yuting Gao, Jiurun Chen+4 more|arXiv (Cornell University)|Jan 19, 2021

Advanced X-ray Imaging Techniques55 references34 citations

TL;DR

This paper proposes Color Transfer Fourier Ptychographic Microscopy (CFPM), a high-throughput full-color digital pathology method that transfers color texture from a low-resolution (LR) full-color image to a high-resolution (HR) grayscale FPM image captured using a single wavelength. By leveraging unsupervised color transfer in Lab color space, CFPM reduces acquisition time by 2/3 with only a 0.4% increase in RMSE (5.7% vs. 5.3%) compared to conventional tri-wavelength FPM, enabling fast, accurate, and efficient full-color imaging compatible with advanced FPM techniques.

ABSTRACT

Full-color imaging is significant in digital pathology. Compared with a grayscale image or a pseudo-color image that only contains the contrast information, it can identify and detect the target object better with color texture information. Fourier ptychographic microscopy (FPM) is a high-throughput computational imaging technique that breaks the tradeoff between high resolution (HR) and large field-of-view (FOV), which eliminates the artifacts of scanning and stitching in digital pathology and improves its imaging efficiency. However, the conventional full-color digital pathology based on FPM is still time-consuming due to the repeated experiments with tri-wavelengths. A color transfer FPM approach, termed CFPM was reported. The color texture information of a low resolution (LR) full-color pathologic image is directly transferred to the HR grayscale FPM image captured by only a single wavelength. The color space of FPM based on the standard CIE-XYZ color model and display based on the standard RGB (sRGB) color space were established. Different FPM colorization schemes were analyzed and compared with thirty different biological samples. The average root-mean-square error (RMSE) of the conventional method and CFPM compared with the ground truth is 5.3% and 5.7%, respectively. Therefore, the acquisition time is significantly reduced by 2/3 with the sacrifice of precision of only 0.4%. And CFPM method is also compatible with advanced fast FPM approaches to reduce computation time further.

Motivation & Objective

• To address the time-consuming nature of conventional full-color Fourier ptychographic microscopy (FPM) that requires repeated imaging at three wavelengths.
• To develop a faster, efficient, and accurate method for full-color digital pathology without compromising image quality.
• To establish a color space model based on CIE-XYZ and sRGB for consistent color representation in FPM.
• To evaluate the feasibility and performance of color transfer as an alternative to multi-wavelength FPM for high-throughput pathology imaging.

Proposed method

• Transfer color texture from a low-resolution (LR) full-color donor image to a high-resolution (HR) grayscale FPM reconstruction using unsupervised color transfer.
• Convert the donor image from sRGB to Lab color space to decouple luminance (L) from chrominance (a, b) for robust color transfer.
• Perform color transfer by matching the cumulative distribution functions (CDFs) of the chrominance channels (a, b) between the donor and acceptor images.
• Use a single-wavelength FPM reconstruction as the acceptor image, selected based on optimal dye absorption to maximize contrast.
• Establish a color space model based on the CIE-XYZ standard and map to sRGB for display compatibility.
• Validate the method across 30 biological samples using quantitative metrics such as RMSE and recovery time.

Experimental results

Research questions

• RQ1 Can color texture from a low-resolution full-color image be effectively transferred to a high-resolution grayscale FPM image to achieve full-color reconstruction?
• RQ2 How does the performance of the proposed CFPM method compare to conventional tri-wavelength FPM and wavelength-multiplexed FPM in terms of accuracy and acquisition time?
• RQ3 What is the impact of illumination wavelength selection on CFPM performance, and how does it relate to dye absorption characteristics?
• RQ4 Can CFPM maintain high image quality while reducing acquisition time by two-thirds compared to conventional methods?
• RQ5 Is the CFPM method compatible with advanced fast FPM techniques to further reduce computation time?

Key findings

• The average root-mean-square error (RMSE) of the conventional tri-wavelength FPM method is 5.3%, while CFPM achieves 5.7% when compared to ground truth, indicating minimal accuracy loss.
• CFPM reduces acquisition time by approximately 2/3 compared to conventional FPM, as only a single wavelength is used for FPM reconstruction.
• The method is compatible with fast FPM approaches, enabling further reduction in computation time.
• The performance of CFPM is wavelength-dependent, with optimal results achieved when the illumination wavelength matches the absorption peak of the stain used.
• CFPM performs best for samples stained with a single dye; it may fail to reproduce correct colors when multiple dyes with distinct absorption profiles are used.
• The average RMSE for the WMFPM method is 11.85%, which is significantly higher than both conventional FPM and CFPM, indicating its limited accuracy despite faster acquisition.