- Challenges of Artificial Intelligence Adoption in Healthcare
KLAS used the terms machine learning (ML) and natural language processing (NLP) when collecting research to validate AI in imaging. Machine learning was defined as tools to learn computer systems algorithms and statistical models to effectively perform tasks without instructions. Natural language processing was defined as enabling software solutions to understand, process, and analyze interactions between computer and human languages.
IBM Watson Health was the most popular vendor in 2018 with 24 mentions. Many experts believed that this vendor was best positioned to deliver AI in imaging. But since then, mentions involving the vendor have declined significantly, only receiving about 15 mentions last year.
Zebra Medical Vision, GE Healthcare, and Nuance followed behind IBM Watson Health. Their customers were impressed with efforts to ensure customer satisfaction and neutrality, as well as the vendors’ access to capital data and optimism.
Although iSchemaView was mentioned by just three respondents, they received the highest average confidence rating. People praised the vendor’s RAPID technology and its delivery quickness.
Many vendors who received under 5 mentions significantly grew in recognition in 2019.
Aidoc and Nuance saw the biggest increases in mindshare. Aidoc was described as being “ahead of the game” and “having a fast-moving target approach.” Although Nuance customers have not yet seen concrete AI deliverables, many respondents believed the vendor’s current speech recognition base may ensure an advantage in AI imaging.
Although not specifically mentioned, niche/start-up vendors were noted to be flooding the AI market and were reported as “innovative but potentially lacking long-term viability.”
In 2018, many interviewed organizations felt more time and research were needed to understand the importance of AI in imaging. Now, almost half (46 percent) of organizations said that they are currently adopting AI, 38 percent answered that they have plans in the works, and 16 percent have no current plans.
No organizations in 2019 reported that they are more than five years away from deploying AI in their healthcare systems. Twenty-nine percent believe they are less than a year away, 53 percent said they are one to two years from deployment, and 18 percent are three to five years away.
These statistics vary greatly from 2018 when only 21 percent of organizations saw deployment less than a year from then. And 17 percent believed their organizations were five or more years away from AI adoption.
Organizations today are implementing deeper and more involved AI systems. This includes more use cases and more imaging departments than before. Of those organizations currently implementing AI, 48 percent answered that they will have a high depth of deep AI adoption. Following close behind, 43 percent of organizations saw a medium depth of adoption, and a modest eight percent saw a low depth of adoption in their future.
“A year or two ago, my radiologist and cardiology thought that AI signified doomsday. Today, those same physicians are asking when they can get their hands on AI. They see the value that AI can bring to them and their patients,” a radiology VP explained in the report.
Of those organizations currently making AI adoption plans, a majority (61 percent) expect a medium adoption, while only 18 percent see a high depth of adoption. Twenty-one percent of respondents reported low adoption in their future.
Many provider organizations place great weight on the references of their colleagues when making technology purchase decisions. The following vendors were top-mentioned in high mindshare for AI imaging:
- Agfa Healthcare
- Adioc
- GE Healthcare
- IBM Watson Health
- Nuance
- Philips
- Sectra
- Siemens
- TeraRecon
- Zebra Medical Vision
By 