ANN-PSO Hybrid ML-Optimization of a Hollow-Disk Resonator-Based Photonic Crystal Optical Sensor for HeLa Cell Tumor...

Area of Science:

• Photonics
• Biosensing
• Nanotechnology

Background:

• Cervical cancer detection requires sensitive and early diagnostic tools.
• Photonic crystal sensors offer label-free detection based on refractive index changes.
• Conventional design methods for photonic crystals face limitations in optimization.

Purpose of the Study:

• To propose a novel two-dimensional photonic crystal optical sensor for early cervical cancer detection.
• To develop an artificial intelligence framework for optimizing the sensor's geometric parameters.
• To achieve high sensitivity and a low detection limit for biosamples.

Main Methods:

• Utilized a two-dimensional photonic crystal with a central hollow-disk micro-cavity and GaAs rods.
• Employed a Random Forest algorithm to explore the design space.
• Developed a multilayer artificial neural network (ANN-MLP) as a surrogate model, coupled with Particle Swarm Optimization (PSO).

Main Results:

• Achieved an average sensitivity of 5512.91 nm/RIU.
• Obtained a quality factor of 6139.15.
• Reached a detection limit of 5.64×10-5 RIU with high tolerance to fabrication deviations (±20 nm).

Conclusions:

• The AI-assisted design strategy effectively optimizes photonic crystal optical sensors.
• The proposed sensor architecture demonstrates superior performance metrics for biosensing applications.
• This approach offers a promising pathway for early cervical cancer diagnosis.