Restoring Brain Tissue Images with Algorithms, Without Expensive Equipment

The long-standing belief that "high-end equipment is essential for obtaining clearer images inside living organisms" has been broken. A new technology has been developed that restores blurry images from inside the brain with remarkable clarity-all without the need for expensive optical measurement equipment. This breakthrough is enabled by a physics-based artificial intelligence (AI) computational algorithm.

Professor Ki-Sung Kang, KAIST

원본보기 아이콘

On April 21, KAIST announced that a research team led by Professor Kisung Kang from the School of Electrical Engineering at KAIST and Professor Na Ji from UC Berkeley has developed a technology that precisely corrects distortions in microscope images used to observe the inside of living organisms, using neural field models.

A neural field model is a neural network-based technique that continuously represents the structure of three-dimensional space, allowing for simultaneous restoration of both images and shapes.

The microscope used in this research is a "two-photon fluorescence microscope." This device uses two weak light sources simultaneously to selectively illuminate only specific points deep within living tissue, allowing researchers to peer inside biological structures.

However, as light travels through thick tissue, it bends and scatters, resulting in distortions. This is similar to how objects appear warped and blurry underwater. This phenomenon is known as optical aberration, where the distortion of light causes the focal point to become blurred.

Traditionally, expensive hardware such as a wavefront sensor-which measures how much the light bends-was essential for correcting this light distortion.

A Framework for Fluorescence Microscope Distortion Correction. KAIST

원본보기 아이콘

In contrast, the joint research team developed an algorithm that calculates and corrects how the light has been distorted by using only pre-acquired image data. Just as software like Photoshop can restore blurry photos to their original form, this method revives clear images without additional equipment.

The core of the technology developed by the team is a machine learning algorithm based on neural field models. This algorithm tracks the distortion process that occurs as light moves, enabling an integrated approach that simultaneously corrects optical aberrations in biological tissue, minute movements within the living organism, and even mechanical errors in the microscope.

Using this technology, the research team successfully obtained high-resolution and high-contrast images from deep within living tissue without any additional optical measurement or correction equipment. This software (algorithm)-based solution challenges the conventional wisdom that expensive equipment is necessary for quality imaging.

This development is expected to lower the barrier of acquiring costly research equipment, enabling more researchers to perform precise brain observations.

Professor Kang stated, "This research is significant in that it opens the way to more accurately visualize the inside of living organisms by combining optical and AI technologies. The joint research team plans to further develop this technology into an 'intelligent optical imaging system' in which microscopes autonomously find optimal images."

The results of the study were published on April 13 in the life sciences methodology journal "Nature Methods."

© The Asia Business Daily(www.asiae.co.kr). All rights reserved.