A novel approach for foreign substances detection in injection using clustering and frame difference.

Area of Science:

• Pharmaceutical Technology
• Machine Vision
• Quality Control

Background:

• Ensuring the purity of injectable medications is critical for patient safety.
• Traditional methods for detecting foreign substances in injections can be time-consuming and may lack sensitivity.
• The presence of visible particulate matter in parenteral products is a significant quality concern.

Purpose of the Study:

• To develop and validate a novel machine vision-based technique for the automated detection of foreign substances in injectable ampoules.
• To classify and effectively detect two types of foreign substances: slowly subsiding and rapidly subsiding objects.
• To improve the efficiency and accuracy of foreign substance inspection in pharmaceutical manufacturing.

Main Methods:

• A machine vision system utilizing controlled spin/stop movements of injection ampoules to induce relative motion of foreign substances.
• Image preprocessing techniques including noise reduction and motion detection.
• Two distinct detection algorithms: Moving-object Clustering (MC) for slowly subsiding objects and Frame Difference for rapidly subsiding objects.

Main Results:

• The proposed machine vision technique demonstrated effective detection of visible foreign substances in 200 tested injection ampoule samples.
• Moving-object Clustering successfully identified slowly subsiding contaminants based on invariant features.
• Frame Difference accurately detected rapidly subsiding contaminants by analyzing frame-to-frame changes.

Conclusions:

• The developed machine vision approach offers an effective and automated solution for detecting foreign substances in injections.
• The dual-method strategy (MC and Frame Difference) allows for robust classification and detection of different contaminant types.
• This technique holds potential for enhancing quality control processes in pharmaceutical production.