Enhancing Cellular Microscopy: Artificial Intelligence-Powered Fluorescence Feature Compensation and Multi-Planar Reconstruction via Cycle-Generated Adversarial Network
Abstract
In the field of molecular diagnostics, Fluorescence in situ hybridization (FISH) is pivotal, notably influencing blood analysis, early tumor detection, and genetic studies. Despite its significance, FISH faces the technical challenge of uneven staining and high noise, which impede the precise delineation of chromosomal features in cell nuclei. Conventional microscopy, although adept at identifying chromosomal aberrations, struggles with the manual intensity adjustments required when processing voluminous datasets, leading to potential data loss or distortion. Addressing this, this study introduces an approach driven by artificial intelligence (AI) – a fluorescence feature compensation method utilizing a Cycle-Generated Adversarial Network (Cycle-GAN). This AI contribution lies in the automated enhancement of fluorescent features through advanced grayscale mapping, which, when paired with the calculated weightings of channel and spatial images, distinctly accentuates the desired features. On the engineering application front, we employ this AI framework for the meticulous layer-by-layer compensation of intracellular fluorescence, culminating in a multi-planar reconstruction of images. Our method not only augments to the average intensity of the three fluorescence characteristics increased by 30.65%, 33.91%, and 48.42%, respectively, but also demonstrates a 36% average improvement in fluorescent spot feature visibility compared to traditional imaging techniques. The application of this enhanced visibility through multi-planar reconstruction expands the depth and continuity of chromosomal features, thereby providing a more robust dataset for the investigation of cellular spatial arrangements. This advancement offers a significant leap in the accuracy and depth of molecular diagnostics, contributing valuable insights into the spatial distribution of intracellular elements, which is crucial for unraveling cancer mechanisms and crafting targeted therapeutic approaches. Our research bridges the gap between AI’s potential in image processing and its practical engineering application, setting a new standard for precision in FISH analysis
Submission
Submited to Engineering Applications of Artificial Intelligence at 01/03/2024.
Citation
Lemin Shi, Mingye Li, Zeng Li, Wei Wang, Ren Xu, Ping Gong, Dianxin Song and Xin Feng. Enhancing Cellular Microscopy: Artificial Intelligence-Powered Fluorescence Feature Compensation and Multi-Planar Reconstruction via Cycle-Generated Adversarial Network. Engineering Applications of Artificial Intelligence. 