Decoding orbital angular momentum in turbid tissue-like scattering medium with deep learning | JoVE Visualize

Area of Science:

Optics and Photonics
Machine Learning
Information Science

Background:

Structured light beams with orbital angular momentum (OAM) are vital for high-capacity optical communications and advanced biomedical imaging.
Multiple scattering in turbid media significantly distorts OAM beam properties, hindering topological charge retrieval.
Accurate OAM decoding is crucial for maintaining signal integrity and image quality in scattering environments.

Purpose of the Study:

To evaluate the effectiveness of deep learning models in classifying the topological charge of OAM beams after undergoing multiple scattering.
To assess the performance of different deep learning architectures, including a CNN baseline, AFT-CNN, and ResNet18, for OAM decoding.
To determine the limits of OAM information survival and retrieval in low-scattering regimes using three-channel measurements.

Main Methods:

Experimentally acquired three-channel intensity and interference measurements from 25 independent acquisition sessions.
Trained and evaluated a Convolutional Neural Network (CNN) baseline, an Angular Fourier Transform CNN (AFT-CNN), and a pretrained ResNet18 model.
Classified signed 11-class and unsigned 6-class topological charges under varying scattering conditions.

Main Results:

The best-performing models (CNN and ResNet18) achieved high accuracy (near 95%) in the low-scattering regime ([Formula: see text]).
Model accuracy dropped sharply at higher scattering levels ([Formula: see text]).
Sign-dependent OAM information was found to survive multiple scattering in the low-scattering regime.

Conclusions:

Deep learning models can successfully decode sign-dependent OAM information from structured light beams even when subjected to multiple scattering.
Three-channel measurements combined with deep learning offer a viable approach for OAM decoding in low-scattering turbid media.
These findings have significant implications for improving the robustness of optical communication systems and biomedical imaging techniques operating in scattering environments.