Digital Color Microscopy Imaging With White-Light ML Reconstruction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing digital microscopy techniques face limitations in achieving high resolution and color imaging efficiently, with methods like bright-field microscopy being time-consuming and requiring multiple image captures for color representation, while Fourier ptychographic microscopy (FPM) is slow due to the need for narrow-band light sources and lacks precision in color representation.
Innovation Solution
Employing a machine learning model trained with a training set of digital images captured using white light emitting diodes and varying illumination patterns to construct a high-resolution color image, leveraging information from multiple angles and refractive indices to improve imaging efficiency and accuracy.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Loss of information
If multiple sequences of images are captured using different colored LEDs to achieve color imaging, then color information is obtained, but the imaging time increases significantly
Solution Approach 1:
The patent combines multiple narrow-band LED illuminations (red, green, blue) into a single white light illumination source. The machine learning model processes a single image captured under white light to simultaneously extract both color information and high-resolution details, eliminating the need to capture multiple separate image sequences with different colored LEDs.
Solution Approach 2:
The patent introduces a machine learning model as an intermediary that processes the single image captured under white light illumination. This model reconstructs the color image by predicting color information from the intensity image, thereby bridging the gap between single-image capture and multi-color-information retrieval without requiring multiple sequential captures.
2Measurement precision
If high magnification is used to improve screening precision, then cell identification accuracy increases, but the visible field of view decreases
Solution Approach 1:
The patent transitions from a single image capture to capturing multiple images at different illumination angles and positions. By synthesizing these multiple low-magnification images with the machine learning model, the system achieves high-resolution detail (equivalent to high magnification) while maintaining a wide field of view (equivalent to low magnification), effectively adding the dimension of angular diversity to resolve the contradiction.
Solution Approach 2:
The patent divides the sample into multiple regions and captures images of each region from different illumination angles. The machine learning model then synthesizes these segmented images into a single high-resolution color image of the entire sample, allowing comprehensive coverage without sacrificing detail resolution.
3Manufacturing precision
If narrow-band LED sources are used for Fourier ptychographic microscopy, then high resolution is achieved, but color precision is lost
Solution Approach 1:
The patent changes the illumination parameter from narrow-band (single wavelength) to white light (broad spectrum). By using white light LEDs that emit across the visible spectrum, the system captures color information directly in the intensity image, which the machine learning model then processes to produce accurate color images while maintaining high resolution through angular diversity synthesis.
Applied Scientific Principles
This section explains which scientific principles are used to turn an abstract innovation direction into a practical engineering solution.
Function Achieved in This Case
The method allows for rapid construction of high-resolution color images with improved precision and broader field of view, reducing the need for multiple image captures and enhancing spectral information capture compared to conventional methods.
Implementation Method 1
The trained machine learning model is then used to construct a digital color image of a sample from a set of digital images of the sample. The plurality of digital images may then be used to determine a refractive index associated with the sample, or to determine phase information associated with the sample.
Implementation Method 2
The plurality of digital images may then be used to determine a refractive index associated with the sample
Data Source
AI summary
The present inventive concept relates to a method and a device for training a machine learning model to construct a digital color image depicting a sample. The method comprising: acquiring a training set of digital images of a training sample by: illuminating, by a plurality of white light emitting diodes, the training sample with a plurality of illumination patterns, and capturing, for each illumination pattern of the plurality of illumination patterns, a digital image of the training sample; receiving a ground truth comprising a high-resolution digital color image of the training sample, wherein a resolution of the high-resolution digital color image is relatively higher than a resolution of at least one digital image of the training set of digital images; and training the machine learning model to construct the digital color image depicting a sample using the training set of digital images and the ground truth. The present inventive concept further relates to a microscope system and a method for constructing a digital color image depicting a sample.