- Imaging analytics aided by artificial intelligence is rapidly expanding within the health IT sphere, but providers must assess their organizations and understand exactly how this technology will enhance the patient experience and reduce healthcare costs before making any investments.
According to a recent report from ReportBuyer, the global medical imaging analytics market will increase to $4.26 billion by 2025 as providers adopt tools to extract meaningful insights from individual images as well as larger-scale data assets.
The rising demand for analytics software reflects the industry’s shift to value-based care. A recent Frost & Sullivan analysis states that the US medical imaging market is becoming an industry driven by quality rather than quantity, with stakeholders increasingly focused on improving safety, cutting costs, and enhancing workflow efficiency.
However, completing this transformation is no easy task. Implementing new infrastructure is costly, and it is often difficult to know whether new tools will produce a return on investment.
Medical imaging tests account for $10 billion of Medicare spending annually, and repeated medical imaging tests substantially contribute to annual healthcare costs, according to a study published in the Journal of the American College of Radiology.
A 2016 study from RAND Corporation added that imaging for back pain and headaches were among the top-performed low-value services in healthcare, costing the industry $3.1 million and $3.6 million each year, respectively.
"The effect of both internal and external challenges has already taken its toll on the industry. The way for stakeholders to not only address these challenges but also succeed will be to judiciously use products, services, and solutions that improve the efficiency of the imaging process, reduce costs, and improve efficiency without compromising quality," said Tanvir Jaikishen, Senior Research Analyst at Frost & Sullivan.
To make the switch from quantity to quality in medical imaging, healthcare providers must consider what technological solutions will most accurately analyze images while reducing costs.
Artificial intelligence, particularly deep learning, has shown significant promise in improving image analytics. Deep learning, a machine learning technique, mimics the decision-making logic of the human brain and offers an increasingly accurate way to classify texts, images, and other clinical data.
Developers and researchers have been showcasing these capabilities through pilots, studies, and use cases that are producing remarkable results.
For example, researchers at Case Western Reserve University developed a deep learning network that could identify the presence of invasive forms of breast cancer in pathology images with 100 percent accuracy, performing better than human pathologists.
Recent research from Google also demonstrated that imaging analytics fueled by machine learning algorithms could identify metastasized breast cancer rates with improved accuracy over other automated processes and pathologists.
“Our method yields state-of-the-art sensitivity on the challenging task of detecting small tumors in gigapixel pathology slides, reducing the false negative rate to a quarter of a pathologist and less than half of the previous best result,” the Google researchers said.
“Our method could improve accuracy and consistency of evaluating breast cancer cases, and potentially improve patient outcomes.”
While advancements in technology like AI and deep learning certainly hold promise for the future of imaging analytics, providers must consider which methodologies will best fit their needs and goals before investing in cutting-edge tools.
The rapidly growing medical imaging industry, combined with the ongoing shift to quality-based care, allows vast opportunities for technology vendors to make a profit, and many are quick to sell their products to healthcare providers who want to improve their organizations.
"The US medical imaging industry is in the process of transforming itself to address the challenges of the present and the demands of the future. This transformation provides unprecedented opportunities for market participants to address the challenges of care providers and develop new products and solutions that help make the imaging workflow more productive," said Jaikishen.
It might be easy for providers to get carried away with excitement and expectations over the potential of AI tools, but they must remember that this technology won’t magically change the way healthcare is delivered.
Despite vendor enthusiasm, AI is still very much in its infancy, and providers should think of these tools as clinical support capabilities rather than machines that can make medical decisions completely on their own.
In addition, organizations shouldn’t jump into agreements with vendors before knowing exactly what they’re offering, and more importantly, how the technology will be used to solve problems.
“When we first started talking to our vendor, the CEO never spent more than two or three sentences describing the technology,” Sameer Badlani, MD, FACP, CHIO at Sutter Health recently told HealthITAnalytics.com. “That wasn’t the selling point. His pitch was solving the problem, and I think that often gets missed in the hype about AI and machine learning.”
Providers should also assess where their organization stands in terms of data integrity and clinical workflows before deploying new health IT tools. Organizations that create a detailed roadmap outlining their major goals will ensure that their technological investments will have meaningful results.
Avoiding “black box” AI tools should also be a critical concern for organizations. When using AI to make clinical decisions, providers must make sure that they understand why and how these systems are making particular recommendations and associations, even if it is difficult for clinicians to understand exactly how an algorithm functions.
AI may be able to uncover actionable insights and assist clinical decision making, but a major challenge will be the ability of AI algorithms to function effectively on complex data streams, according to a recent JASON report produced on behalf of the ONC.
The report states that AI is currently both powered and limited by its access to digital data. Health data is frequently locked away in siloes, making it difficult to combine with external sources. Additionally, organizations are often discouraged from sharing data assets due to privacy concerns, so it is challenging for AI to make the most of their algorithms.
Providers also must consider how big a role AI will play in the future of their organizations. While studies show that deep learning tools can potentially outperform human clinicians in extracting results from medical images, this doesn’t mean AI should completely replace radiologists.
A recent study examined the ability of deep learning networks to predict the duration of kidney function in patients with chronic kidney disease and found that deep learning tools can act as a supplement to clinicians.
Although the calculations of the deep learning algorithms were more precise than those of traditional pathologist-estimated scoring systems, the researchers stated that the traditional approach is still highly valuable. A pathologist’s clinical impression takes other contextual factors into account, while deep learning algorithms only visually inspect a part of the body in isolation.
While there is no doubt that imaging analytics aided by AI tools like deep learning will be a significant part of the future of healthcare, providers should assess their organizations before making any investments and understand that this technology can’t magically fix their problems.
To improve diagnosis accuracy, enhance patient care, and cut healthcare costs, providers must understand exactly how AI tools will assist with care delivery.