Electronic Proofreading: Robots and Vision in the Pharmaceutical Industry

The importance of providing quality products for all manufacturers is a top priority; for the pharmaceutical industry, it is absolutely critical. At the top of this list is incoming material inspection and proofreading. Pharmaceutical manufacturers have good reasons for concern as, over the years, these areas have been the largest reasons for recalls in the industry. 

To help handle these challenges, the Food and Drug Administration (FDA) adopted new guidelines for manufacturing practices to prevent mix-ups on the production line; unfortunately, they left out incoming inspection. A single hyphen missing from the label copy changes the directions for usage from 'take 1-2 tablets daily' to 'take 1 2 tablets a day'. Such a potentially fatal error must be picked up by the pharmaceutical companies' incoming quality assurance department. The critical nature of this task puts a tremendous burden on the incoming inspection departments. As a result, they are now required to inspect incoming lots of thousands, if not hundreds of thousands, of labels, cartons, inserts, and containers to prevent 'minute' errors.

Proofreading
To meet FDA Good Manufacturing Practices (GMP) guidelines, every pharmaceutical manufacturer has established departmental procedures to make sure incoming materials, such as labels, inserts, and cartons, carry information that is exactly the same as what the manufacturer originally approved. Often called a 'golden' or master copy, a signed original is kept on file to compare against newly printed incoming materials. But rather than making life easier, these masters add to the proofreader's burden, as many master 'proofs' are not the same as the incoming material. Some are photocopies, laser prints, or color keys. If you add in thousands of SKU's and products, plus exporting in different languages, it becomes an almost impossible task. Tylenol, for example, has over 70 different SKU's alone.  While high speed, online inspection is improving thanks to the use of bar codes and computer vision systems, incoming inspection of text and graphic copy is still often done by one or two inspectors comparing 'master' to 'sample.'

Astra-Zeneca
Some companies go even further. In addition to the Good Manufacturing Practices that drug companies must follow, Astra-Zeneca, has added a set of Military Standards for passing or failing inspection of an incoming lot of materials. The Military Standard, in this case Rule 105-E, establishes guidelines for inspection which Astra-Zeneca applies to its labels, cartons, inserts, and containers as follows. The company starts with a sample lot of 800 as 100% inspection of 150,000 inserts is not feasible or reasonable. Defects, for example on an insert with text and graphics, are divided into different critical levels for rejecting the entire lot of 150,000.

These include:
-non-Tolerance: i.e. incorrect name, product, or bar code. Any one sample will reject the entire lot.
-critical: i.e. unreadable legal copy or blank areas of legal copy. 0.65% of the 800 samples will cause a reject.
-majors: i.e. color out of specification; torn, wrinkled, or blurred copy. 1.0% of the 800 samples will cause a reject.
-minor Defects: i.e. hickeys, off-register, missing (non-critical) print, misspellings. 4.0% will cause lot to be rejected.

Automated vision improves accuracy, reduces inspection time by 70%
To handle these kinds of inspection challenges, in the early 1990's GSMA's sister company, Parish Automation, Inc., introduced the first electronic image comparator (proofreader) for the pharmaceutical industry. Called Inspec, the system works as follows:

The operator would place a label, leaflet, or carton under the CCD camera and capture an image in memory using a Coreco Imaging vision board, a predecessor to Coreco Imaging's current PCVision high performance PCI bus frame grabber. The image is sent to a computer processor which creates an electronic negative. This is then used to subtract the incoming sample from the master. Any differences are highlighted and 'flashed' on the television monitor for the operator to view.

The real challenge, according to Gary Parish, is that 'many proofreaders are under the impression that they are as accurate and faster than electronic inspection. The best way I found to demonstrate the problems human inspectors face was to ask the proofreader to take two drawings from a 'Find the 10 Differences' comic strip and compare his performance with that of the computer. The proofreader usually takes several minutes, if they are even able to find the differences, while the machine vision-based computer can highlight the differences in less than one second'.

Materials come in all sizes, so the manual approach requires an operator to either step the camera or material through each area of comparison. The area of inspection or 'Field of View' (FOV) of the CCD camera is based on the size and complexity of the material being inspected. If an operator is inspecting a small label with Japanese copy, a small 'FOV' may be required to pick up a missing character stroke. This area could be less than 1' x 1' of the material. Since the CCD camera is based on television technology, the image is made up of pixels (picture elements), usually 659 H x 494 V. The greater the number of pixels in a character, the easier it is to pick up any differences.

Adding robotics
To take income materials inspection to the next level, the addition of a fully automated Cartesian x-y robot allows the camera to be stepped through the entire insert for inspection at the resolution required.

The next generation of the Inspec system are GSMA's RoboSpect and PageSpect. These vision systems combine robotics with machine vision technology to reduce inspection times by up to 95%. RobotSpect automatically inspects press sheets up to 40' x 60' for printers to the pharmaceutical industry. PageSpect is designed as a table-top version for materials up to 12' x 18' in size, as well as roll or pin-fed labels.

To meet strict validation requirements, both systems are designed to control the entire inspection process. In addition to Coreco Imaging's PCVision frame grabber, which captures and transfers images to the host processor in less than 4 ms, GSMA has added:

• Fully programmable x-y robotic inspection stage,
• Single or multiple cameras with different resolution,
• Controlled lighting, enclosed workstation,
• Locked-in camera parameters,
• Windows 95 operating system, and
• High speed computer with ability to store masters in memory.

Both systems provide a number of unique features that are specially designed for pharmaceutical use. An exchangeable inspection stage allows manufacturers to advance and inspect 100% of both pin-fed and roll-fed labels. Unlike high-speed web inspection, pharmaceutical inspection requires the proofreader to determine the level and acceptability of the error therefore both systems have supervision acceptance password levels and automatic stop/subtract functions. Each system also has fail-safe features, noting errors and operators who pass them.

The current CCD versions of RoboSpect and PageSpect can inspect up to 60 labels per minute. An entire 8' x 11' insert can be inspected in less than one minute, depending upon the amount of copy on the page.