Cloud Over Troubled Data: Healing Bio-IoT
By Joe Stanganelli
April 25, 2018 | As the healthcare and life-sciences sector has gradually accepted cloud-based data solutions, and as the sector's Big Data still grows ever Bigger, the Internet of Things (IoT) has become a driving force for health-IT practitioners and clinical researchers. All that data, after all, must come from somewhere—and must be constantly maintained and tempered by even more data.
"You can assume that maybe 90% of the data that's generated for healthcare [will] be generated outside of a hospital or outside of a medical setting," Paimun Amini, Ag Productivity IT Lead, (R&D and Commercial) at Monsanto Company, recently told Bio-IT World, pointing to the proliferation of consumer health wearables, agricultural biotic sensors, and "smart" peripherals for electronic laboratory notebooks (ELNs). "So IoT is playing a huge aspect from the perspective of how it transforms how we're going to get the data and begin making insights."
And yet, when you ask David Iyoha, Director of Software Solutions at systems-integration firm Fortech, what the most common pain point is for healthcare and life-science IoT applications, he tells you that it's the data.
It's confusing. It's unreliable. It's inconsistent. It can't be trusted.
"One of the big challenges—that, many times, are not considered—is [making] sure that the data that we're getting from the devices or sensors are always good," Iyoha says. "You want to be very or fairly confident that the data you're getting from all these devices are actually accurate and believable and usable."
Considering the ubiquity and diversity of IoT data aggregators, this is no small task.
Automating Human Error
Within the process of automated IoT data collection, explained Iyoha, there are numerous hazards. Upgrades, reboots, and other preventative maintenance can add unanticipated complexities. And, of course, there's human error.
"As time goes on, defects get introduced into the system," said Iyoha. "So you need to have processes… that can identify that there is a defect, track the defect down, and get to the root cause—and then put in fixes, document those fixes, and then systemize that."
Systemization, however, can be taken too far. Philip Skinner, a product manager at PerkinElmer, described to Bio-IT World how major pharmaceutical companies spent the past 15 years or so investing tremendous resources in extensive automation—heavily configuring ELNs, such that each organization "ended up with a bespoke ELN of their own." Consequently, in the race to the smart lab of tomorrow, standardization seems to have fallen by the wayside as a matter of proprietary self-defeat.
"This is [supposed to be] a process…to ideate and consolidate on datasets," said Skinner. "The ELN was essentially a developer environment—a developer toolkit."
"The rapid growth of the IoT in general would not be possible without the growing availability and maturity of cloud services," Jeff Kaplan, Managing Director of cloud-computing and managed-services consultancy THINKstrategies, told Bio-IT World, "and [neither would] specific advancements in the life-sciences/healthcare sector in particular."
Kaplan and other cloud-computing experts have long held that IoT and the cloud go hand in hand—that you cannot effectively have the former without the latter. Skinner is no exception when it comes to clinical IoT, arguing that the problems of on-premises ELNs and other on-premises smart-lab tools are twofold: (1) that updated on-prem ELNs and smart-lab tools are difficult to migrate and distribute, and (2) that the datasets themselves from these tools face similar migration difficulties because of ontological inconsistencies.
As a product manager for a web-based ELN solution, Skinner acknowledged sua sponte the self-serving nature of his position, but maintained that "these on-prem solutions get really old and dated." At the same time, Skinner prophesies that the combination of IoT and cloud laboratory-as-a-service solutions could evolve to the point of overpowering the ELN as we know it today.
"I think the beauty of cloud-based solutions as a whole is [that] you can innovate…and you can start to push the boundaries of preconceived ideas—like you can get rid of the [electronic lab] notebook," said Skinner. "At some point, we could just phase it out, and at the same time we can push out new things very quickly."
To some extent, these things have already begun to happen.
Enabling Monsanto's Mesh and Multicloud IoT
Climate FieldView—an initiative by Monsanto subsidiary The Climate Corporation that Amini describes as "probably Monsanto's biggest IoT play"—purports to collect and combine weather data, elevation data, soil conditions, and other pertinent information on farming conditions that allows farmers to anticipate fertility, water uptake, disease emergence, crop damage, flooding, and the like.
"What farmers care about [is that] they 100% want to do real-time monitoring of the conditions on a farm, on the field," explained Amini. "And that's when you begin thinking about things like mesh networks—where if I have 20 different sensors spread out over a couple of miles here, can they be connected, [and] can they be streaming the data together so I can have more of an overall picture of the area?"
This kind of networked interconnectivity among smart sensors becomes even more crucial in the clinical laboratory setting because of the globalized nature of clinical collaboration and scientific reproducibility.
Monsanto relies heavily on a multicloud implementation for its smart lab solutions. "Every single lab that we work with has different aspects of mass scales or scanning systems for 1D- or 2D-bar codes or RFIDs. To enable that, a lot of it is linked via IoT solutions that we implement in the lab [in] real time," said Amini. "As you have more instruments [connected to a cloud network], you have an opportunity to begin standardizing some of this data collection and automating it into a way that we don't have to worry about this process as much anymore."
"Really, you are actually minimizing human error," Amini added.
Indeed, for Monsanto's global seed-processing and molecular-breeding pipeline in particular—where shipping, processing, and genomic sequencing data and insights have to gathered and acted upon rapidly—this kind of automation yields the consistency and quality-control benefits with which David Iyoha is so concerned.
"[W]e have to make very quick decisions and turnaround times by automating a lot of these steps via some of our IoT solutions that are taking and processing this information there," explained Amini. "Our labs are not that unique from others."
Lab IoT: A Recipe for Reproducibility
Little surprise, then, that lab-equipment vendors are acting upon these demands for agile consistency. Earlier this year, sample-management manufacturer Gilson announced the simultaneous release of "smart" liquid-handling (i.e., pipetting) tools with its own cloud platform for pipetting. While Gilson's "Internet of Pipetting" solutions are designed to work with open-source ELN sciNote, Gilson seems to tout that the ELN aspect here—beneficial as it is to auditing and human reinterpretation—is incidental to the real pain point the company seeks to solve with its smart-pipetting suite: the problem of reproducibility.
"90% of scientists agree that reproducibility is in a crisis in science right now," Gilson spokesperson Erika Dix told Bio-IT World in an email statement, referencing a 2016 Nature survey of 1,576 researchers. Between this statement and a related press release, Gilson avers that measurement flaws stemming from improper pipetting technique combined with data-recording and reporting issues—including "missing notebooks, deleted emails, and corrupted files"—lead to irreproducibility. Smart lab instruments operating on a common platform-as-a-service, the company claims, achieve the reliability and consistency that have been lacking in the world of the traditional, siloed, non-standardized ELNs such as Skinner describes.
"Data's dirty," said Skinner. "So enabling the standards—or enabling semantic understanding to retrospectively apply standards onto the dirty data you get in the course of [science]—will really help bring data together very elegantly. That's the goal."
Editor’s Note: Paimun (PJ) Amini and David Iyoha are both speaking at IoT workshops at the upcoming Bio-IT World Conference & Expo in Boston, May 15-17. For more details see D2K, Transforming Data to Knowledge with Cloud, IoT & Machine Learning (AI) – Part I (Amini) and the Bio-IT IoT Workshop (Iyoha).