Application of artificial neural networks for Process Analytical Technology-based dissolution testing.

Area of Science:

• Pharmaceutical Sciences
• Analytical Chemistry
• Computational Modeling

Background:

• Pharmaceutical tablet dissolution is critical for drug efficacy.
• Process Analytical Technology (PAT) enables real-time monitoring of manufacturing processes.
• Predictive modeling of dissolution from PAT data is essential for quality control.

Purpose of the Study:

• To apply Artificial Neural Networks (ANN) for non-destructive prediction of in vitro tablet dissolution.
• To evaluate ANN performance against Partial Least Square (PLS) regression using Process Analytical Technology (PAT) data.
• To explore the utility of ANN for data fusion in PAT applications for pharmaceutical manufacturing.

Main Methods:

• Utilized Near-Infrared (NIR) and Raman spectroscopy to collect spectral data from extended-release tablets.
• Developed and compared Partial Least Square (PLS) regression and Artificial Neural Network (ANN) models.
• Modeled tablet dissolution directly from spectroscopic data, considering tablet composition and compression force.

Main Results:

• ANN models demonstrated superior predictive accuracy, achieving up to 3% lower root mean square error of prediction (RMSEP) than PLS models.
• The ANN model using reflection NIR spectra yielded the most accurate dissolution predictions (mean f1=6.5, f2=63).
• ANN effectively served as a data fusion method without extensive preprocessing.

Conclusions:

• ANN provides a powerful tool for non-destructive prediction of pharmaceutical tablet dissolution from PAT data.
• The proposed method enhances real-time release testing and improves the efficacy of dissolution testing.
• This approach can significantly advance data processing within the PAT framework for pharmaceutical quality assurance.