Hybrid AI models for thermal imaging and analysis of neurological disorders using thermoplasmonics

Abstract

This paper describes a novel Artificial Intelligence-based approach to predict and analyse heat distribution in multi-tiered tissue structures by using plasmonic nanoparticle enhanced Multi-Spectral Thermal Imaging (MSTI). The optimization problem combines biophysical simulation with innovative machine learning techniques to improve the thermal mapping and analysis of biological tissues.

The described technique uses gold (Au) and silver (Ag) nanoparticles of sizes 25–35 nm, being characteristic of their thermoplasmonic properties and capable of obtaining high-resolution thermal images through multi-spectral imaging. A new Rank Entropy Machine Learning (RE-ML), incorporating probabilistic hidden chain features and entropy analysis of the thermal patterns resulting from plasmonic nanoparticle interaction, is presented.

The RE-ML framework then regenerates thermal distributions which undergo global and local entropy characterizations assessment before a probabilistic Hidden chain model's feature ranking determines the features to be preferentially used. The system obtains 97.8 % accuracy in specific tissue-level pattern recognition, excelling in neurological tissues; high precision of 98.6 %; sensitivity of 98.5 %; and specificity of 99.0 % in visualizing and analyzing thermal distributions over different tissue regions.

Despite its 0.14 error rate, the system is highly accurate in forecasting heat distribution. Thus, the proposed approach exhibits unmatched precision in thermal pattern recognition and presents possibilities for enhancing the heat maps of biological tissues.