This story has been corrected to note that Copan Diagnostics' WASPLab offers two-way track integration. We regret the error.
NEW YORK – Automation has transformed clinical laboratory disciplines like chemistry and hematology, but has until recently left the microbiology space relatively untouched.
However, a combination of rising sample volumes, cost pressures, and staffing shortages, along with technological advances has led to increasing uptake of automation by clinical microbiologists. And while the field has traditionally relied heavily on manual processes, at this point almost every step in a typical clinical bacteriology workflow can be automated.
Additionally, several firms are developing AI-based approaches to reading culture plates that could further streamline the process.
The move to automation "is really taking place as a result of the challenges around the accessible supply of qualified technicians and scientists," said David Newsome, global marketing leading, lab automation, at Becton Dickinson, which is one of the major players in automating the clinical microbiology space.
Consolidation of microbiology labs is upping the sample volumes individual labs have to process while at the same time reducing the number of facilities available to train new lab staff.
"On top of that, there's also the rise of antimicrobial resistance, which is increasing the complexity [of clinical microbiology]," Newsome said. "So automation is a solution to a lot of those challenges."
BD's Kiestra Total Lab Automation (TLA) system and Copan Diagnostics WASPLab system are the two most comprehensive automation platforms, with both covering essentially the entire microbiology workflow and offering integration into two-way track systems. A number of other companies, including Clever Culture Systems, i2A Diagnostics, and NTE Healthcare, offer partial automation systems covering various portions of the clinical microbiology workflow.
Roughly speaking, microbiology automation systems can be broken down into three parts: specimen processors, which take samples arriving to the lab and streak them onto culture plates; incubators, where cultures are grown and monitored via digital imaging; and automated colony pickers, which take the bacterial colonies selected by the microbiologist as requiring following up and transferring them to, for instance, a MALDI plate for MALDI-TOF-based identification, or using them to prepare suspension fluids for antibiotic susceptibility testing.
These components replicate manual workflows that typically consist of a half-dozen or more touchpoints for lab staff. In the absence of automation, samples first have to be sorted and batched according to what kind of test will be run on them. Technicians then culture specimens on media plates that have to be handled and sorted, then incubated, and then sorted again based on what amount of growth is observed.
In a manual system, technicians have to go in and out of the incubator to check plates for growth, which disturbs growing conditions, leading to slower and less consistent growth. Additionally, plates are typically checked at a single point during a shift. This means that plates that went into the incubator late in one shift probably won't have incubated for long enough when the tech comes to check for growth and may have to wait until the following shift to come out.
Allowing techs to monitor plate growth remotely using digital imaging limits disturbances of the incubation conditions while also providing more flexible timing around when to check plates — both factors that streamline the culturing process.
Gilbert Greub is the director of the Institute of Microbiology at the University Hospital of Lausanne, where he said his lab can see as many as 800 samples per day. The facility installed a BD Kiestra system in 2017. He said that installation of the full microbiology automation system had saved his lab the equivalent of about two full-time employees, though he noted that due to rising test volumes his lab had not actually cut any staff but had, rather, used automation to keep up with the increasing demand.
Automation has had other advantages for his lab, he said. "You get, for instance, less contamination. You have less risk of error, increased traceability, and reduced time to results."
A complete automation system like the Kiestra is a substantial investment, noted Greub, who authored a 2016 review in Clinical Microbiology and Infection weighing the pros and cons of different automation platforms. He declined to say what the University Hospital of Lausanne paid for their system but said that they typically run between €1 million ($1.1 million) and €2 million.
Newsome declined to say how many microbiology automation systems BD Kiestra has sold, but according to the CMI review authored by Greub, as of August 2015, the company had placed 105 specimen processing units and 68 TLA systems in 173 labs globally. According to the review, Copan had as of that time placed 325 specimen processing units and 23 lab automation systems (with another 11 installations pending) in a total of 243 labs. Silvio Lignarolo, senior business development manager at Copan, said last month that the company was approaching 700 systems worldwide.
Lignarolo also suggested that despite the pressures driving increased automation in clinical microbiology, the space was unlikely to see major new entrants. "The market has become, I don't want to say saturated — there is still a lot of potential for growth," he said. "But I don't see this being a market where a lot of big players will come in, given the size of the market and that it's already, I think, at a pretty mature stage."
Michael Mitchell, director of microbiology services at the University of Massachusetts Memorial Medical Center, which is served by the Quest lab, said that while the lab hasn't done a formal evaluation of the cost or time savings provided by the system, it is clear from his experience, thus far, that it takes fewer people to process more samples.
A key emerging direction for automating the microbiology lab is implementation of AI-based image analysis to aid technicians in assessing plate growth. A number of companies, including BD, Copan, Clever Culture Systems, and i2A Diagnostics are active in this area.
Matthew Mackechnie, microbiology laboratory supervisor at Quest's Marlborough lab said, in fact, that the AI tools provided by their Copan system are among its most useful components.
"This software is what has actually made the [automation] line really successful for us and is where we are gaining a lot of efficiency," he said, noting that currently the lab uses the software to sort likely positive plates from likely negatives prior to review by a technician.
"If I were, say, looking at throat cultures, this software would go in and put the positives in one folder for me and the negatives in another," he said. "Now I'm in the mindset of looking at the negative cultures just to make sure they are correct and then really quickly releasing the results. All the time it would [normally] take to get them sorted [into positives or negatives] has been gained because this software has put it into a folder and it's now ready for me to make that call."
In the future, the lab plans to use the software for more advanced analyses, Mackechnie said.
"The most exciting thing is that it has the ability to look for things that we define," he said. "Say we want it to look for positive urine cultures from a particular type of organism. We can define the software to do that. We haven't deployed it to do that, but that is the potential it has."
Clever Culture Systems, which is based in Switzerland and is a joint venture of Australia's LBT Innovations and Germany's Hettich Holding Beteiligungs- und Verwaltungs, this year received US Food and Drug Administration 510(k) clearance for its APAS Independence instrument, which automatically screens, interprets, and sorts urine culture plates.
Greub said he believed that in the future automation-based image analysis might be used not just to assess whether a culture is positive or negative but to also make microbial identifications, potentially supplanting existing technologies like MALDI-TOF mass spectrometry.
"We are working with specialists in image analysis that might enable automated identification without MALDI-TOF," he said. In 2017, Greub and collaborators at BD published a proof-of-concept study in Biomedical Journal investigating how effectively image analysis software could quantify growth and identify microbes, and found that the software could identify microbes with accuracy ranging from 98.3 percent to 99.7 percent.
BD currently offers an AI-based tool for analyzing urine cultures that Newsome said "can assess the amount of growth on a plate… and provide an early detection flag to potentially accelerate the time to result."
He added that the company is "actively exploring new AI-based applications for microbiology laboratories to assist in plate reading."
Copan's Lignarolo said that he also saw expanding the capabilities of the company's AI-based tools as a key area of focus for the future.
He noted that the complexity of the microbiology sample typesprovided substantial challenges in this regard.
"It's one thing to apply AI to a simple microbiology test like, say, a urine culture, but another to apply that technology to something more complex like a wound culture or a stool sample," he said. "In [the latter] case, you need a more complex set of rules and algorithms."
"That is definitely one of the main areas where we are working to improve and increase the value of our solution," he said.