Image-Based Artificial Intelligence Methods for Product Control of Tablet Coating Quality.

Area of Science:

• Pharmaceutical Manufacturing
• Image Analysis
• Quality Control

Background:

• Pharmaceutical production relies heavily on visual inspection and experience for process control.
• Objective quality assessment of coated tablets is crucial for ensuring product efficacy and safety.

Purpose of the Study:

• To investigate the efficacy of image analysis techniques for classifying the quality of coated tablets.
• To compare the performance of different classification algorithms in this application.

Main Methods:

• Coated tablets with varying coating levels were imaged using a conventional office scanner.
• Image segmentation was employed to extract numerical data from individual tablet images.
• Four classification techniques were applied: Support Vector Machine (SVM), Convolutional Neural Network (CNN), Partial Least Squares (PLS) regression, and a numerical threshold model.

Main Results:

• The Support Vector Machine (SVM) demonstrated superior performance in terms of computational time and classification accuracy compared to the Convolutional Neural Network (CNN).
• Partial Least Squares (PLS) regression was the fastest method but yielded lower classification accuracy.
• A numerical threshold classification model achieved accuracy comparable to the SVM approach.

Conclusions:

• Image analysis provides a viable and objective method for classifying coated tablet quality.
• Multiple classification techniques, including SVM and numerical threshold models, are effective for this purpose, offering flexibility in implementation based on specific needs.