AI-Powered Microscopy Turns Optical Flaw into High-Quality Imaging Breakthrough

Researchers have harnessed generative AI and chromatic aberration to achieve high-quality, label-free quantitative phase imaging with a single exposure—paving the way for more accessible and efficient biomedical diagnostics.

Research: Single Exposure Quantitative Phase Imaging with a Conventional Microscope using Diffusion Models

*Important notice: arXiv publishes preliminary scientific reports that are not peer-reviewed and, therefore, should not be regarded as definitive, used to guide development decisions, or treated as established information in the field of artificial intelligence research.

Quantitative phase imaging (QPI) is a microscopy technique widely used to investigate cells and tissues. Even though the first biomedical applications based on QPI have been developed, both acquisition speed and image quality must improve to guarantee a widespread reception.

Scientists from the Görlitz-based Center for Advanced Systems Understanding (CASUS) at Helmholtz-Zentrum Dresden-Rossendorf (HZDR), Imperial College London, and University College London suggest leveraging an optical phenomenon called chromatic aberration—which usually degrades image quality—to produce suitable images with standard microscopes.

Employing a generative AI model, just one exposure is needed to obtain the image quality necessary to make QPI attractive for applications in biomedicine. The team presented the work in late February at the 39th Annual Conference on AI by the Association for the Advancement of AI (AAAI), which was organized this year in Philadelphia (USA). The corresponding peer-reviewed conference paper will be available later in March.

Labeling biological samples with dyes or other agents reveals valuable insights. However, this approach has some disadvantages that hinder its widespread usage in clinical diagnostics: It is time-consuming, and expensive equipment and reagents are needed. Research in the past years has, therefore, centered on certain label-free microscopy methods like QPI. Here, the magnitude of light absorbed from or scattered by the sample is not the only one of interest. Using the scattering information, QPI also captures how the sample shifts the phase of light passing through it – a change directly related to its thickness, refractive index, and other structural properties. While also QPI requires quite expensive equipment, computational QPI does not.

One of the most prominent computational QPI approaches is solving the Transport-of-Intensity Equation (TIE). This differential equation allows for calculating an image of the sample based on the recorded phase changes. The approach is easy to integrate into an existing optical microscope set-up and results in good-quality images.

On the downside, the TIE method often requires multiple acquisitions with different focus distances to eliminate artifacts. Dealing with through-focus stacks can be time-consuming and technically demanding, so this type of TIE-based QPI is often not feasible in a clinical setting.

Making use of chromatic aberration

"Our approach relies on the similar principles as TIE but only needs one image because of a clever combination of physics and generative AI," says Prof. Artur Yakimovich, Leader of a CASUS Young Investigator Group and corresponding author of the work presented at the AAAI Conference.

Information about the phase shift induced by biological specimens does not come from additional exposures taken with other focus distances. Thanks to chromatic aberration, a through-focus stack can also be generated from one single exposure.

Most lens systems of the microscope cannot perfectly bring all wavelengths of (polychromatic) white light to a single converging point – a handicap that only highly specialized lenses can correct. This means, e.g., red, green, and blue (RGB) light have slightly different focus distances. "By recording the phase shifts of those three wavelengths separately using a conventional RGB detector, one can build a through-focus stack that facilitates computational QPI, turning the handicap into an asset," Yakimovich explains.

"Using chromatic aberrations to realize QPI poses one challenge: The distance between the red light focus and the blue light focus is minimal," says Gabriel della Maggiora, PhD student at CASUS and one of the two lead authors of the publication. Solving the TIE the standard way just does not give meaningful results. "We then reasoned that we could use artificial intelligence. As it turned out, this idea proved to be decisive", della Maggiora adds. "After training a generative AI model with an open-access data set consisting of 1.2 million images, the model was able to retrieve phase information even though just relying on the minimal data input from the recording."

Method validated on real-world clinical specimen

The team drew on a generative AI model for image quality improvement presented last spring: the Conditional Variational Diffusion Model (CVDM). It belongs to a particular family of generative AI models named diffusion models.

The developers emphasize that training a CVDM requires significantly less computational effort than training other diffusion models, while the results are the same or even better. Harnessing a CVDM strategy, della Maggiora and colleagues developed a novel diffusion model applicable to quantitative data. With this model, they could finally realize computational QPI based on chromatic aberrations.

They validated their generative AI-based approach using, for example, a common brightfield microscope equipped with a commercially available color camera to make microscopic images from real-world clinical specimens. Analyzing red blood cells in a human urine sample, the method unveiled the donut-like shape of these cells, whereas another established computational TIE-based approach was not.

The virtual absence of cloud artifacts in the images calculated with the new generative AI-based quantitative phase imaging variant was an additional advantage.

The Yakimovich group "Machine Learning for Infection and Disease" develops novel computational techniques for microscopy that could be immediately applied in clinical settings.

The potential, e.g., in diagnostics, is enormous. Among the methods used is generative AI. As generative AI is prone to producing hallucinations, the group's primary focus is to reduce them. Incorporating physics-based elements is the key approach here. As the AI-based quantitative phase imaging example shows, this approach is very promising.

*Important notice: arXiv publishes preliminary scientific reports that are not peer-reviewed and, therefore, should not be regarded as definitive, used to guide development decisions, or treated as established information in the field of artificial intelligence research.