To See or Not to See?

Visual Inspection remains one of the key challenges for pharmaceutical manufacturers. It remains a key challenge for inspections and operations because legal requirements are often not easily translated into company policies and strategies.

In the world of visual inspection, there are many questions related to particles that are frequently asked:

• What is a visible particle?
• How to inspect for it?
• What is acceptable?
• How to assess deviations or particle findings?
• What does that mean for the batch status?
• How to qualify operators?
• What is innovative approach for defect detections and inspection?
• How to validate automated visual inspection?
• How to leverage artificial intelligence (AI)?

Patient Safety First

In recent years, visual inspection has become an increasingly important aspect of pharmaceutical production. It is now considered one of the fundamental elements of high-quality injectable production to ensure the safety of medicines. It is no longer just a means of detecting and eliminating defective elements produced upstream.

For this reason, the last revisions of pharmacopeias have further elaborated on visual inspection. Assessing the potential risk to patient safety is a key component to consider.

Manual Visual Inspection

Once the knowledge of the production process is established, the visual inspection capabilities to verify the performances, especially regarding various defect types, need to be established. Critical defects are the ones that could be a potential risk to patient safety.

The first and most powerful visual instrument available is the eye, or better, the eye-brain system. It is an incredibly powerful tool that can achieve very high performances, but every sophisticated tool needs much maintenance to achieve stable and accountable results. Training visual inspection operators based on a defect library to transfer the process knowledge is paramount. Our eye-brain system is very good at looking for what it has been instructed to look for, and it is flexible enough not to be biased by context variability with reasonable accuracy. When assessing visual inspection operators’ capabilities, the industry can use Knapp sets with defects of different relevance: size, contrast, shape, color, etc. Different kinds of contaminants are important when establishing the respective sensitivity thresholds.

One may realize that visual inspection is not a single-person band but a holistic approach and needs the collaboration between development, operation, quality control, packaging, and vision expertise that is also coordinated by quality assurance and regulatory responsibility. Overall, visual inspection is a team effort.

Automated Visual Inspection

Once visual inspection is assessed and has sufficiently large quantities of units and batches, we can transfer the knowledge to automated equipment.

Actual regulation considers the equipment to mimic the visual inspection capabilities while providing reproducibility due to an automated approach. However, they lack the flexibility to be resilient to natural variations in context. This is the main source of false rejection in a well-tuned automatic visual inspection (AVI) machine.

Even in this case, the human expert is at the center of the automatic inspection process. The individual has to be trained in process knowledge before developing recipes and fine-tuning the quality level required to secure production. Several powerful tools are available to ease the individual’s job, like storing images and playing back with optimization strategies for fine-tuning. However, the human expert’s role, which includes sensitivity and critical thinking, is key to achieving the best results in predicting potential anomalies.

Again, to assess AVI performances, we can utilize Knapp tests, possibly with higher abundance and variability, to cover the production process in the best way possible. Acceptable quality limit (AQL) sampling will then continuously verify that the inspection process is well-established. For this reason, we still need a well-established manual visual inspection team to check AQL samples for AVI.

The last revisions of pharmacopeias have stressed the importance of assessing trends of visual inspection to detect potential adverse trends in process quality and drive continuous improvements. In this role, we also need visual inspection experts who can classify known defects and recognize new ones to trigger an improvement process involving the whole production and quality team.

Is Artificial Intelligence the Next Big Thing?

In previous years, the main goal of AVI was to increase inspection process efficiency and productivity while (at least) comparing well to the quality of the visual inspection.

AI has the potential to overcome these limitations by combining the high-speed performances of classical AVI and the flexibility and knowledge-based approach typical of machine learning. As mentioned, the role of the human visual expert is paramount in feeding the AI system with correct and relevant classified examples and verifying AI’s statistical consistency of the knowledge base used to train the machine learning models.

Certainly, this is a new exciting chapter and journey in visual inspection that could further improve production safety and provide advance tools for deepening production knowledge.

Conclusion

Visual Inspection is a never-ending journey to quality and safety for the patients and an incredibly powerful tool for process knowledge. Such a wonderful journey could not be successful without involving all the small and medium-sized enterprises responsible for the production process. The latest technology will empower new capabilities if we are able to embed it in the holistic approach.

PDA’s flagship event, the 2024 Visual Inspection Forum, is back on April 9-10 in Munich, Germany, where the patient is at the center of our vision. We hope to visually detect you in Munich!