Page 21 - OHS, October 2022

work alongside each other in the same space. It could be that the future distribution center has humans paired with biomechanical “exoskeletons” working alongside autonomous forkliftfts. In order for this to work, AI is needed to plan smart paths, create collision avoidance and perform advanced biometric modeling to name a few. AI is also being used to run elaborate simulations that can prepare response plans for catastrophic events. Imagine that extreme weather has damaged a chemical plant, or an explosion has occurred in a complex mining operation. Machine learning models can help with adaptive changes to ventilation systems, evacuation routing and more.
Another advancement—and perhaps how technology will change the workplace the most in the next decade—is the tools that are being built that change how workers perform the everyday work that they do. For example, all over the world, clinical labs are experiencing a shortage of qualified lab analysts. A lab analyst’s job is difficult, tedious and both physically and mentally exhausting. Additionally, special training is required to perform the necessary tasks at a high level. These jobs also tend to typically be done by workers who are aging out of the workforce, and employers are having a hard time replacing them when retirement comes. Pair this with a growing aging population, and you have a perfect storm of a sharp increase in the need for more lab tests with a decrease of qualified workers to perform the analysis. Labs are turning to AI-driven diagnostics with human support to find sustainable solutions. This process involves digitizing work and using various forms of AI to assist humans.
First, samples are prepared and then scanned by a digital microscope. Once the slide is digitized, it can be analyzed by AI. In this stage, the AI can be trained to look for specific objects of interest, for example, white blood cells, parasites, cancerous cells, bacteria and many others. Next, the AI counts and classifies those objects to either label the slide as positive or negative. A human can then confirm the predictions made by the AI. In some cases, this technology can find hard-to-locate pathological organisms more often than humans, resulting in more accurate diagnoses. In others, it can reduce the time it takes for technicians to analyze a slide from three to five minutes to under 30 seconds3 per slide while improving accuracy and consistency.
In a recent study to compare the results of manual microscopy and AI-driven diagnostics, digital diagnostics were five times more sensitive.4 A similar study showed that digital diagnostics found two additional positive cases that had been missed by manual microscopy out of 135 positive results by technicians.4 As one can imagine, the accuracy and sensitivity demonstrated by this type of technology can impact lab productivity along with patient care and health across the globe.
This AI-driven diagnostics technology isn’t just being used for human healthcare. There are new AI-based analyses that industrial hygienists and environmental professionals are utilizing. AI is being used to help perform analyses on spore traps for mold and pollen, dust characterization and characterizing nanomaterial emissions. For spore trap analysis, software companies are taking existing laboratory methodologies that have been around for decades and actually improving them. The inconsistency of spore trap analysis is no secret in the indoor air quality space. In order to make these analyses economical for commercial laboratories, most laboratory SOPs only analyze about 20 percent to 30 percent
of the sample and then extrapolate the data to get an estimated total for the entire sample. Since the particulate loading in a spore trap is not uniform, this can lead to huge variability in sample results. AI can analyze an entire spore trap sample in a fraction of the time it takes a human. Since the whole sample trace is analyzed, this results in more reliable and more repeatable data in an industry that has been stagnant with progress in microscopic analyses.
It is important to realize that AI is not perfect, and just like implementing any new process or system in the workplace, a detailed evaluation should be performed. An AI-based system can only be as good as the data it is built on, and even with careful consideration, there can be biases in the data. Imagine if, when building and testing a collision detection system on a forklift, they had only used workers wearing orange hi-vis clothing. If this hypothetical system was in place in a workplace where workers were wearing green hi-vis clothing, it is possible that this bias could mean the system would not detect when the forklift is about to contact a worker wearing green. Great care needs to be taken by the people developing these solutions to make sure biases are not inadvertently introduced into the product. Another concern with implementing AI in the workplace is privacy and data security. Workers may hesitate to take needed breaks or use the restroom if they feel that their every move is being monitored by management, which could have health and performance implications, so care is needed as more of these solutions come to market.
Workers may have an initial fear of AI encroaching on their workspace. The thought that robots are coming to take our jobs has existed in science fiction for decades, but the reality is that artificial intelligence for most industries will come in the form of tools that will make some aspects of our jobs less burdensome. By taking on some of the process-oriented tasks that computers are well equipped to handle, it will leave humans with more time to perform more value-added tasks or tasks that they are better equipped to do.
The first Industrial Revolution took a generation to transform how we view work in this country, and this “fourth Industrial Revolution” has just started to hit its stride. Synchronous work between humans and technology will continue to change how we live and work. I feel that together, the potential for a better workplace increases. How will it impact our lives next?
Dylan McIntosh is a Certified Industrial Hygienist (CIH) who has performed hundreds of mold assessments in residential, commercial, and healthcare settings. He is also a PAACB Certified Spore Analyst who has analyzed thousands of airborne and surface mold samples. Dylan is currently the product and data manager for Sporecyte, the leading AI platform for fungal analysis.
REFERENCES
1. https://www.osha.gov/data/commonstats
2. https://www.bls.gov/news.release/cfoi.t02.htm
3. https://journals.asm.org/doi/10.1128/JCM.02053-19
4. https://techcyte.com/press-events/study-confirms-the-sensitivity-of- techcytes-ai-solution-for-intestinal-protozoa-detection/
www.ohsonline.com
OCTOBER 2022 | Occupational Health & Safety 17