Trending AI and Digital Solutions in Radiology—current adoption, applications and continuing developments

GE Healthcare

Doctor in virtual space looking at AI medical exams

In the wake of the global COVID-19 pandemic, the use of artificial intelligence (AI) in healthcare is on the rise. Progress in anything seemed to be held static while leaders and healthcare providers focused on stopping the spread of COVID-19 and treating infected patients, but the opposite was true in AI. The pandemic, in fact, amplified the need for the continued development and integration of digital solutions and AI into healthcare and brought about tangible solutions.

Doctors and patients readily adopted new methods of care delivery using technology, hospitals needed to provide ways for clinicians such as radiologists to have access to patient imaging files and work remotely, and AI solutions were quickly employed in imaging exams to assist in COVID-19 diagnoses. Many healthcare leaders and clinicians are interested in keeping the tangible benefits that resulted from the use of AI-enabled tools during the pandemic.

According to a survey of healthcare leadership by Optum, 56 percent said they are accelerating or expanding their AI deployment timelines in response to COVID-19, demonstrating the importance of this business tool during the most stressful times[1]. Many radiologists already familiar with radiology AI applications are looking forward to more progress in AI and additional AI applications. As adoption in radiology increases and AI-enabled procedures and applications are approved for reimbursement, many more will be brought to market. Independent AI developers as well as industry leaders such as GE Healthcare are working together to bring valuable AI innovations to the forefront of radiology to streamline workflow and impact health outcomes.

Advantages of AI for radiology

While it is important that health leaders continue the trajectory in incorporating AI tools into hospitals across a wide variety of applications, AI tools are particularly well suited for applications in radiology. AI algorithms can help with streamlining workflow by automating repetitive tasks, for example, but because radiology is a data-driven environment dependent on visual interpretation, AI-enabled software and solutions hold significant potential in clinical applications. Currently, many AI solutions have been deployed across all major medical imaging modalities, enabled by deep learning through which complex tasks, such as identifying an abnormality in a CT scan can be performed quickly and accurately.

As the AI market for radiology continues to grow, many radiologists and other clinicians are eager to understand and work with the new AI enabled solutions and digital tools that could enable them to work more efficiently, allow more time for and facilitate collaboration with other members of the patient’s care team, and ultimately, to impact diagnostics and treatment.

In an IMV report on Global Imaging Department Priorities and Outlook, radiologists reported their top departmental priorities as improving department workflow efficiency and productivity, improving patient satisfaction, and keeping their department up to date with state-of-the-art technology[2]. Many AI tools are available to help them support their efforts in these areas.

Some AI tools focus on increasing efficiencies in the scheduling and technologists’ workflows. Predictive modeling can be used to avert no show appointment situations that can have a negative impact on scheduling, for example. Other AI tools can be used to reduce errors or automate tasks. Some of the most effective AI tools can seem very simple conceptually yet have a significant workflow impact. An AI tool created by GE Healthcare for X-Ray can automatically rotate the image on the screen for the technologist during the exam, saving three to four mouse clicks from the process with every patient, which can add up to 70,000 clicks per year for a single technologist[3]. Now extrapolate that time savings to the estimated 4.2 billion imaging procedures that were performed in 2019, of which X-ray and ultrasound tests made up slightly more than 80 percent[4].

AI is also helping clinicians acquire images faster in ultrasound, speeding up exam times which has the potential to positively impact patient experience and satisfaction. Taking methodical assessments of heart function, for example, can be tedious and time-consuming to acquire. High quality data acquisition and operator skill are key elements needed to achieve accurate and complete exams. But these patients also typically need subsequent monitoring exams, making consistency and reproducibility vital in identifying disease improvement or progression. New AI-enabled ultrasound technology decreases operator variability because it can semi-automatically detect the appropriate measurement of spectral Doppler images, enabling the system to fast-forward the path from scanning to measurements and streamlines the scanning process to reduce scan time without sacrificing accuracy.

Clinically, many of today’s advanced imaging technologies utilize AI-based image reconstruction. In CT, for example, GE Healthcare’s TrueFidelityTM AI uses deep learning neural networks to intelligently suppress noise from CT image data without suppressing the anatomical structure data. The technology can be used in a wide range of clinical applications and gives an image appearance^ similar to traditional FBP images while maintaining the performance of ASiR‐V in the following areas: image noise (pixel standard deviation), low contrast detectability, high‐contrast spatial resolution, and streak artifact suppression*.

The world market for AI-based medical imaging products and services like the above examples, according to an IMV report, was estimated at $300 million in 2019 and is expected to quadruple by 2024 to $1.2 billion[5]. Deployed across multiple imaging modalities, nearly 75 percent of these products focus on just four clinical areas: cardiology, breast health, pulmonology, and neurology. According to the report, increased new product activity is expected in these clinical areas as well as expanding to other areas such as for the detection of kidney, liver, ophthalmic and orthopedic abnormalities. Progress may be further fueled by the steady rise in the number of AI enhanced medical imaging procedures that are approved for third-party reimbursement.

Poised for the future together—AI and radiology

Many new AI solutions are in the pipeline targeted for healthcare, with a reported $4 billion invested into the sector in 2019, up from $2.7 billion in 2018[i]. Brought to market by a variety of developers, from start-ups to original equipment manufacturers (OEMs), new AI solutions show tremendous promise and the potential to alleviate some of the challenges faced by radiologists.

With so many developers in the market, and the market ready to adopt the technology, it appears to be a win-win situation. However, some established players in the industry that already have regulatory clearances, active users and recurring revenues represent serious challengers to newcomers. OEM developers can not only put more resources behind their AI development, but they also have deep domain expertise within their installed base, working with the technology and most importantly, with the clinicians.

Working cooperatively is one way that great AI solutions can reach clinicians. As an industry leader, GE Healthcare has a proven track record in generating scalable solutions and efficiencies. To deploy its AI solutions, it created the Edison™ Ecosystem. However, GE Healthcare took its Edison concept a giant step further to support, enable and deploy rapid innovations created by independent developers. The Edison™ Developer Program was designed to connect the best innovations in AI with healthcare providers in an intelligently efficient, and scalable way.

GE Healthcare is also a founding partner in the National Consortium of Intelligent Medical Imaging (NCIMI), an academic, clinical and industry collaboration established by the University of Oxford. The first of its kind collaboration includes a network of the UK’s National Health System (NHS) hospitals, and a variety of commercial and industry partners like GE Healthcare and small start-up AI development companies. By bringing together a wide range of expertise into one sustainable ecosystem, NCIMI aims to address unmet needs in AI and to deliver manageable and tangible healthcare transformation in diagnostics, therapeutics and monitoring.

Enterprise AI—radiology’s new frontier

The pressures in healthcare to reduce costs while improving patient experience and health outcomes have only increased over time. Utilizing advanced imaging technologies requires significant financial investment and, like any other capital investment, is tied to a plan for increasing patient volume, revenues, or potentially an expansion of imaging services, to project the investment’s return (ROI).

Incorporating AI tools in healthcare and in radiology for the benefit of streamlining workflows and improving health outcomes, of course comes at a cost and health leaders are beginning to expect an ROI from these investments as well. While over half of health executives ranked improving health outcomes or patient experiences as the greatest impact of their AI investment, they also believe they’ll begin to see ROI in as soon as two years[6].

Outside of radiology, larger scale AI solutions and implementations, known as enterprise AI, are proving valuable by incorporating data from disparate sources such as payor, provider, pharmaceutical and population groups in order to provide insights on population health and value-based care[7].

Moving the concept of enterprise AI into radiology, it has the potential to offer clinicians valuable longitudinal information for a patient over time with respect to their disease progression. However, if an AI tool was able to take measurements not only for the clinical needs of today but record the size and location of everything included in the image, it could impact that patient’s health outcomes far into the future, with an even more significant impact on children who are affected by disease. Additionally, imaging data from one patient with multiple scans could be combined with other patients suffering from the same disease or symptoms and potentially offer clinicians new insights on the disease and its incidence and progression.

The future holds great potential for AI in radiology, and for radiologists, who have been on a constant path forward, making the most of clinical and AI tools that support their efforts and sharpen their expertise in early diagnosis and treatment of disease.

 

For more digital Imaging Solutions, please visit GE Healthcare Imaging.

 

 


[3] GE Healthcare data on file

[4] Global Imaging Department Priorities and Outlook 2019, IMV, https://imvinfo.com/

[5] The Artificial Intelligence In Imaging Landscape, 2020, IMV, https://imvinfo.com/

[i] https://www.fiercehealthcare.com/tech/investors-poured-4b-into-healthcare-ai-startups-2019

^as shown on axial NPS plots

*As demonstrated in phantom testing where image noise was evaluated using the uniform section of the Catphan 600 for both head and body. DLIR and ASiR-V reconstructions were performed using the same raw data.