Human-Inspired Automated Visual Inspection

Intoduction

In pharmaceutical production, the integrity of injectable drugs must be ensured throughout the whole drug shelf-life. Regulatory bodies enforce strict standards that require thorough inspection for visible particulates and container defects to ensure patient safety. Together with the well-known USP <1> [1] for injections and implanted drug products, stating that “all products intended for parenteral administration must be visually inspected for the presence of particulate matter” and that “every container in which the contents show evidence of visible particulates must be rejected”, the USP <790> [2] on visible particulates in injections sets the regulatory requirements for the visual inspection of parenteral products. It enforces the demonstration (through 100 % visual inspection) that the batch is “essentially free of visible particulates” before it is released.

Currently, around 50 % of these crucial quality checks are done by manual visual inspection (MVI), which is still considered the “golden standard” (fig. 1). The regulatory requirements for the qualification of a visual inspection process focus therefore on the equivalence between the automated visual inspection (AVI) methodology chosen and the reference MVI one. To mention some of them, the EU Annex 1 [3], with its revised version released in 2022, the USP <1790> [4] and the Chinese Pharmacopeia ChP 2020 volume 3 [5]. All these guidelines state: “Where automated methods of inspection are used, the process should be validated to detect known defects which may impact product quality or safety and be equal to, or better than, manual inspection methods.”(EU Annex 1 – GMP 2022).

The comparison between the AVI and MVI performances is performed by putting in place a Knapp study [6], which represents an essential phase of the AVI qualification.

Semi-automated visual inspection (SAVI) is another alternative to the MVI. However, both these methods are associated with high labor costs, extensive training and qualification requirements, as well as variability due to human factors such as fatigue and bias. Additionally, it is becoming increasingly difficult for pharmaceutical manufacturers to recruit and retain qualified manual inspectors in the long term.

With the rise of high-mix/low-volume manufacturing and drugs produced in small batches, the pharmaceutical industry is facing increasing pressure to improve the efficiency and accuracy of inspection. This article highlights considerable potential for technological advances in this area.

User Requirements

The motivation to optimize and automate production processes is increasing, due to the several challenges of MVI. Like every human-based process, MVI is prone to bias and fatigue, lacks consistency and reproducibility. Moreover, the work performed is slow and monotonous, making the process somehow inefficient from a business perspective.

In addition to the challenges of MVI, the efficient fulfillment of regulatory requirements is a key driver for changing to AVI systems. The re-inspection of gray outputs from an inline AVI platform, such as the inspection of clinical trials or media fills, offers great potential.

The most important requirements for such systems can be summarized as follows:

Simple integration: The system should not be much larger than 2 MVI stations and must fit seamlessly into existing facilities without requiring structural changes or additional media supplies.
Flexible: It must be achievable to process the widest possible range of standard packaging. It should be possible to test both liquid-filled and lyophilized containers with different filling levels.
Change-over simplicity: Switching quickly between different formats and products is crucial for small batches. Therefore, a format change with the corresponding release should be possible in less than 30 min.
Throughput: An inspection speed 3–4 times faster than a human operator would be desirable.
Compliant: The system ensures compliance with regulatory requirements for visual inspection.
Artificial Intelligence (AI)/Deep Learning: AI integration must be possible to boost performance.

Approach to Automation

Robots are an obvious transportation solution for automating visual inspection. If the human movements involved in MVI are analyzed in detail, they can be transferred to a robotic system. The focus here is on mobilizing any visible particles that may be present. To avoid the generation of bubbles, which can hinder visual inspection, robots have degrees of freedom to optimize the manipulation.

Reproducible sample manipulation, gentle handling and smooth product transportation are also important factors that must be given special attention to achieving optimal inspection results. These are then combined with advanced software algorithms for image processing, which can not only use image subtraction to identify moving objects in consecutive images, but also sophisticated track and trace algorithms. When carefully fine-tuned, tracking of single objects can unravel the position as well as the speed of a moving object through a stack of a few hundreds of consecutive images.

On top of that, AI-promptness of the AVI system will ensure detection performance compliant with the requirements, also when the inspection tasks go beyond the capabilities of standard rule-based algorithms. In particular, the use of deep learning is highly beneficial for the analysis of difficult-to-inspect parenterals (DIPs) and for defect classification.

Inspection precision can be increased thanks to human-inspired automated visual inspection systems that not only have the nuanced skills of the human inspectors, but also deliver consistent, reproducible results.

An Innovative Solution

The solution offered by the latest automated systems is groundbreaking (fig. 2).

In addition to integrating advanced robotics and machine vision technologies that mimic the precision and adaptability of human inspectors, the following aspects have been incorporated into the solution:

Product-centered processes are implemented for minimal impact on the drug and the container. Gentle robotic handling helps prevent damage to the container. Human-inspired particle mobilization allows to get rid of spinning.
Smooth product transport and reproducible sample manipulation help to minimize the risk of damage, contamination and incorrect manipulation.
The optical station uses the latest image acquisition and processing components as well as advanced rule-based and deep learning algorithms to increase detection accuracy. The static optical path (e.g. Wilco OptiX™) dispenses from any movement of the optical components and forms the basis for the flexibility to inspect different container sizes without mechanical adjustments of the optical components.
The dimensions of the system are small enough to fit through standard doors and the plug-and-play installation requires only 1 electrical connection. Both aspects contribute significantly to simplifying installation and integration.
Quick format changes and short recipe development times help to minimize downtimes. The good and easy accessibility of the machine from all sides also has a positive effect on the high utilization of the machine.
The GMP-compliant design allows for easy cleaning and fast line clearing, which is essential for maintaining production efficiency. Internal heat monitoring allows for carefree 24-hour operation.

Use Cases

Automated systems for visual inspection are versatile enough to support various use cases, such as:

Commercial large-scale production: Ejection testing saves inline production systems unnecessary additional testing procedures.
Clinical trials and small batch production: Where precision and adaptability are of paramount importance.
Media Fill Inspection automates traditional manual work.

Scalability

A system that can be used in a variety of ways increases utilization and speeds up amortization. If one operator can also manage several systems at the same time, it is possible to use the system economically for many applications over a long period of time.

In addition to being used as a production system for the inspection of small batches, the solution is also suitable as a development system for offline recipe development or for validation runs, thus avoiding downtime of the main production line.

Use during the drug life cycle

Due to the scalability and the use of multiple systems, it can be utilized in the complete life cycle of a drug. Different numbers of systems can be used in the different phases of the life cycle to inspect the planned production volume.

While one system may be sufficient in the clinical production phase and the start-up phase, several equivalent systems can be a cost-effective solution during the expansion phase. During the blockbuster phase, such a system can take over the eject inspection of an inline system perfectly. Once the drug patent has expired, it is possible to switch back to several small batch systems. Finally, at the end of the drug life cycle, only one system may be sufficient to inspect the corresponding production quantity.

Due to the variable use of such systems over a long period of time, the resulting acquisition and maintenance costs are almost negligible. Depending on the production volume, compact automated systems are amortized within a short time and reduce inspection costs by more than half compared to manual inspection.

Advantages of AVI

Automating the visual inspection process for parenteral products offers several advantages, such as higher throughput and capability compared to manual and semi-automated counterparts, as well as reproducibility and consistency of results. Bias and fatigue effects are completely ruled out, and with the engineering of the specific optical stations, a huge range of the defect categories and locations can be precisely targeted. Systems for AVI have on the other side only some drawbacks, mainly related to the poor adaptability to new defect categories without previous adjustment of the optical stations or fine-tuning of the algorithms, and the presence of false rejections. With careful and well-reasoned machine design, both on the hardware and the software side, these issues can be easily overcome.

When focusing on the visual inspection of small batches, the following benefits arise when switching from MVI to AVI systems:

Operational efficiency: Up to 4 times higher throughput than MVI as well as quick changeover and use of the system throughout the drug life cycle increase overall production efficiency.
Quality assurance: Improved and consistent traceable detection capabilities ensure product safety, regulatory compliance and reduce false reject rates.
Economic efficiency: Significant reduction in labor costs as a single operator can manage multiple systems and the number of manual inspectors can be substantially reduced.

These advantages make the initial investment for an AVI system, both in terms of complexity and business perspective, a robust choice.

Conclusion

Manual visual inspection of small batches of pharmaceutical drug products poses various challenges. A change to automated systems is taking place in the pharmaceutical industry. The requirements for such systems are likely to be relatively similar, but adequate solutions have hardly been available on the market to date. By transferring human-inspired movements to robots, the smallest systems can guarantee the highest inspection quality. They deliver consistent results and efficient inspections. The applications are diverse and range from clinical studies to the gray reinspection of inline systems. The simple conversion to other formats and recipes as well as the scalability of such systems will further expand their uses. They are likely to find their way into pharmaceutical production since they are flexible to use and economically very efficient.

References

[1] United States Pharmacopeia (2023). General Chapter, 〈1〉 Injections and Implanted Drug Products (Parenterals) – Product Quality Tests. USP-NF. Rockville, MD: United States Pharmacopeia.

[2] United States Pharmacopeia (2024). General Chapter, 〈790〉 Visible Particulates in Injections. USP-NF. Rockville, MD: United States Pharmacopeia.

[3] ASTM E2587-12 (2023). EU GMP Annex 1 Manufacture of Sterile Medicinal Products. The Rules Governing Medicinal Products in the European Union, Volume 4 EU Guidelines for Good Manufacturing Practice for Medicinal Products for Human and Veterinary Use.

[4] United States Pharmacopeia (2024). General Chapter, 〈1790〉 Visual Inspection of Injections. USP-NF. Rockville, MD: United States Pharmacopeia.

[5] China (2020). Pharmacopoeia of the People’s Republic of China. Beijing: China Medical Science Press.

[6] Knapp, J. Z., & Kushner, H. K. (1980). Generalized methodology for evaluation of parenteral inspection procedures. PDA Journal of Pharmaceutical Science and Technology, 34(1), 14–61.