Machine learning-enhanced temperature sensing of spectrally featureless fluorophores

Area of Science:

• Optical thermometry
• Fluorescence spectroscopy
• Machine learning applications

Background:

• Conventional fluorescence thermometry relies on probes with multiple emission bands or large spectral shifts.
• Spectrally featureless fluorophores with single emission bands and modest shifts are challenging for accurate temperature sensing.
• Existing methods are limited by excitation source fluctuations and photobleaching, and insufficient peak shift data.

Purpose of the Study:

• To optimize machine learning (ML) algorithms for fluorescence-based temperature sensing using spectrally featureless probes.
• To evaluate ML models for predicting temperature from the spectral changes of resveratrone, an organic fluorophore.
• To overcome limitations of traditional thermometry for probes with single emission bands and modest peak shifts.

Main Methods:

• Tested four ML models: multiple linear regression, support vector regression, linear random forest, and multilayer perceptron.
• Utilized resveratrone, a spectrally featureless fluorophore, for fluorescence-based temperature sensing.
• Compared ML model performance against conventional chemometric methods.

Main Results:

• The pyramid-structure multilayer perceptron demonstrated superior performance in capturing temperature-dependent spectral changes.
• This ML model achieved a relative performance index of 170, indicating significantly lower test Mean Absolute Error (MAE).
• The model's accuracy advantage persisted across different experimental conditions (day, concentration, solvent matrix).

Conclusions:

• Machine learning, particularly the pyramid-structure multilayer perceptron, enables accurate temperature prediction from spectrally featureless fluorophores.
• This approach expands the scope of materials usable for fluorescence nanothermometry.
• The findings support potential applications in biologically and biomedically relevant temperature ranges (25-70 °C).