Accurate inspections of parenteral product packaging is critical to protect consumer safety. Parenterals are products are that patients take via an intravenous, intramuscular or subcutaneous injection, entering the blood stream very quickly and are often used in emergency situations. That’s why, these types of products must be perfectly safe and have no characteristics that would compromise their integrity during storage so that they can be injected safely when used.
Appropriate inspection of injectable products is particularly important for the pharmaceutical industry and poses a number of significant challenges that can be addressed in effective and innovative ways, including through the use of next-gen technologies like neural networks.
For many reasons, more and more pharmaceutical companies are choosing plastic for their injectable product packaging. First and foremost, this material offers major advantages over glass (primarily, it won’t break). However, plastic poses significant challenges during packaging inspections, because it is less transparent than glass and its optical characteristics are more variable, making it more difficult to analyse using Visual Inspection methods.
Traditionally Visual Inspection technology uses blob analysis, and is based on video cameras that collect images of the product from various angles, with the product both still and in motion. The images are then sent to software that can identify groups of connected pixels and identify them as “blobs”, that is, uniform regions of the product having unexpected characteristics. By identifying these regions, the machine can automatically classify the product as not meeting the required quality standards, since blobs could indicate the presence of a crack, a foreign body or particulate matter inside the container, an inclusion in the packaging wall, etc.
However, this type of analysis is less effective when used on products in plastic containers, since the characteristics of plastic can lead to certain optical effects, which means that correctly sealed and stored products might be considered damaged (causing false rejects). More serious, plastic’s reduced transparency means that products not having the expected characteristics might be wrongly accepted as compliant, leading to many possible negative consequences in terms of consumer safety, as well as economic and brand impacts on the manufacturer.
So, an entirely different approach is needed to resolve this problem, one that is no longer based on blob analysis only, but that also uses neural networks. By leveraging the power of AI, machine can “learn” from their errors during the training phase, and determine which products are non-compliant and which ones are compliant —even though they do not match the standard—by considering a large number of variables.
The potential of neural networks ensures accurate results even when testing plastic packaging, significantly reducing the number of false rejects and increasing process safety.
Bonfiglioli Engineering has been collaborating for years with university research groups that are developing technologies based on the use of neural networks.
Research into these technologies is advancing in leaps and bounds, working to provide solutions to increasingly specific needs, including those of the pharmaceutical industry that requires reliable tools to safely identify products not meeting quality standards.
These types of solutions are custom designed to match the characteristics of each individual type of packaging to be tested, opening the way to many innovative integrity testing solutions for pharmaceuticals. Bonfiglioli Engineering is already offering these types of solutions to its customers today.
To learn more about Automatic Visual Inspection of parenteral products in plastic packaging and about Bonfiglioli Engineering’s solutions, watch this video featuring our design manager Davide Luisari.