Researchers at the University of Hong Kong have developed a new microscopy technique that reconstructs three-dimensional images from significantly fewer measurements, potentially accelerating biological imaging while reducing light exposure to samples.

The approach, called AIMED, short for Arbitrary Illumination Microscopy with Encoded Depth, combines optical encoding with computational image reconstruction. The system is designed for multiphoton microscopy, a widely used imaging method for studying biological tissues deep below the surface.

Conventional multiphoton microscopy typically captures images layer by layer, requiring repeated scans across different depths. While effective, the process can be slow and expose samples to large amounts of light, limiting its usefulness for observing fast biological events or conducting long-term studies.

The Hong Kong team sought to overcome those constraints by collecting compressed imaging data and reconstructing full 3D volumes using sparse optimization algorithms. Rather than scanning each depth individually, the system excites multiple layers simultaneously and then computationally separates the signals.

Breaking scanning bottlenecks

To achieve this, researchers used a spatial light modulator to split a laser beam into multiple focal points positioned at different depths. The intensity of each focal spot can be adjusted independently, helping compensate for signal loss deeper inside tissue.

The technique also benefits from the nonlinear nature of two-photon and three-photon excitation, which helps suppress interference between imaging planes. This allows the encoded depth information to be recovered more accurately during reconstruction.

The researchers tested AIMED on mouse brain neuronal samples and reported that it could resolve fine structures such as dendrites and axons while operating at a compression ratio of roughly 60%.

According to the team, the method required only one-half to one-third of the optical power normally used per imaging plane. In some test configurations, reconstructed images showed improved contrast compared with conventional scanning approaches.

Faster imaging, less light

The researchers also examined delicate neuronal features known as dendritic spines. They reported that AIMED delivered image quality comparable to, and in some cases better than, traditional sequential scanning methods that rely on higher optical power.

Across compression ratios ranging from 62.5% to 87.5%, reconstructed images maintained a structural similarity index of about 0.95 and a peak signal-to-noise ratio between 41 and 42 decibels, indicating minimal loss of image fidelity.

The team further conducted simulation studies involving up to 47 imaging planes. Those simulations suggested the method could increase volumetric imaging speed by approximately eight times in large-scale imaging tasks.

Unlike many acceleration strategies that require extensive hardware upgrades, AIMED is designed as a relatively simple add-on approach. Researchers say the framework could be adapted to other imaging methods, including confocal microscopy, Raman imaging, and photoacoustic imaging.

The work could be particularly useful for studying sparse biological structures such as neuronal networks and for imaging applications where minimizing photodamage is critical.

The findings were published in the journal Advanced Photonics.