Oblique Illumination QPI with Deep Learning for Fewer Captures
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current quantitative phase imaging (QPI) techniques are limited by their transmissive nature, restricting their use to thin samples and requiring multiple acquisitions to obtain a desired image, which hinders speed and application on thicker or in-vivo samples.
Innovation Solution
A deep learning neural network (DLNN), specifically a generative adversarial network (GAN) like U-Net GAN, is used to generate quantitative phase images from one or two raw captures, reducing the need for multiple acquisitions.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If multiple acquisitions are used to extract quantitative phase in qOBM, then measurement precision is improved, but productivity deteriorates due to the large number of captures required
Solution Approach 1:
The system performs preliminary action by capturing a series of images with different oblique illumination angles in advance, then uses pre-computed point spread functions (PSFs) corresponding to each angle to reconstruct the quantitative phase map. This allows the quantitative phase to be extracted from a single captured image rather than requiring multiple sequential acquisitions during actual measurement, thereby improving imaging speed while maintaining precision.
Solution Approach 2:
The system uses partial action by selecting and applying only the specific PSFs corresponding to the oblique illumination angles actually used in the capture, rather than processing all possible angles. This selective approach reduces computational overhead while maintaining the precision needed for accurate quantitative phase extraction from the captured images.
2Measurement precision
If transmission-based QPI techniques are used, then measurement precision for thin samples is achieved, but adaptability deteriorates for thicker or in-vivo samples
Solution Approach 1:
The system inverts the traditional transmission-based illumination approach by using epi-illumination (illumination from above) with oblique angles. Instead of transmitting light through the sample from below, light is directed obliquely onto the sample surface and the reflected/scattered light is captured. This inversion enables the system to penetrate and image thicker samples and in-vivo tissues that cannot be effectively imaged with conventional transmission methods.
Solution Approach 2:
The system changes the illumination parameter from normal transmission to oblique epi-illumination, and varies the oblique angle as a controllable parameter. By adjusting the illumination angle and using angle-specific PSFs for reconstruction, the system adapts to different sample thicknesses and types, maintaining measurement precision across a broader range of sample conditions including thick and in-vivo samples.
Data Source
AI summary
An exemplary embodiment of the present disclosure provides a quantitative phase imaging method, comprising: imaging a sample to obtain one or more raw captures: inputting the one or more raw captures into a deep learning neural network (DLNN); generating, using the DLNN, a quantitative phase image of the sample based on the one or more raw captures; and outputting the quantitative phase image.