Cost-Benefit Analysis of AI in Radiology
Reading Time: 5 minutes read
Diagnostic Imaging Providers Need to be Diligent in Evaluating New Technologies
Artificial intelligence (AI) in radiology has been generating significant buzz for several years, and will inevitably play a large and vital role in the future of medical imaging. But what is the cost-benefit analysis for current AI applications in radiology?
As with any emerging technology, healthcare facilities need to be diligent in their cost–benefit analysis to determine AI’s true value and ability to deliver desired results in radiology. Even though a solution is named Artificial Intelligence doesn’t mean that it will automatically provide a benefit to the facility or the patient. AI holds much potential but the solution should be tested against the current level of care to make sure it can deliver clear and quantifiable benefits in either cost savings, workflow or clinical outcomes. (1)
At Carestream, we are actively working on AI-based solutions that could provide proven clinician, patient and business benefits with the goal of adding them to our broad portfolio of radiology systems as soon as the technology is tested and ready.
It is important to have hard reference data when evaluating claims about the efficacy of using AI for clinical decisions. Performance numbers from the vendor on algorithms are not enough (these are highly dependent on the particular database used for testing). Peer-reviewed clinical studies should be conducted to compare actual radiologist performance with and without the use of AI technology. It may turn out that the AI software is actually slowing the workflow for the radiologist or, more critically, compromising care of the patient.
Imaging and Workflow Intelligence delivers quantifiable benefit to clinician and patient
If your objective is to improve workflow, you might consider an intelligent imaging approach. Carestream has deep and proven expertise in image processing that enables us to apply “Imaging and Workflow Intelligence” to diagnostic X-ray exams.
From fixed algorithms to linear regression to machine learning and other AI techniques, we utilize a variety of advanced software tools to produce pristine images that assist the radiologist in making an accurate diagnosis. One example, is our Pneumothorax Visualization Software, built on a fixed algorithm that has proven to help radiologists detect the presence of pneumothorax.
A study of 206 portable ICU chest radiographs (103 with pneumothoraces) were processed with and without image enhancement software and reviewed by five readers who varied in reading experience. The mean Area Under the Curve (AUC) for pneumothorax detection increased for 4 of the 5 readers, with the largest improvement for the reader with least experience.
Additionally, our Eclipse image processing engine, which powers Carestream software, offers options that include EVP Plus Image Processing, Tube & Line Visualization, Bone Suppression, SmartGrid, Anatomy Clipping Software and Pneumothorax Visualization. Carestream will continue to add to our suite of Eclipse intelligent imaging and workflow options with increased utilization of AI technology, all with the goal of delivering quantifiable benefits to the clinician and the patient.
AI’s role in radiology for the foreseeable future
Adoption of AI in the healthcare sector has been slow, according to the Advisory Board. Data privacy concerns, regulatory hurdles and interoperability between vendors are among the reasons for this slow adoption.
For these reasons and more, we suggest that AI’s appropriate role, for the foreseeable future, is that of a supplemental lens for medical image analysis. In this capacity, AI holds much potential for improving operational efficiency, freeing up experts from repetitive and mundane tasks, and supplementing the skills of a radiologist by identifying subtler changes in scans while reducing treatment planning time by analyzing vast amounts of data. In this arena, the cost-benefit analysis of AI in radiology is favorable.
Looking farther ahead, we should not expect AI to replace radiology’s human component—as radiologists bring far more value to the diagnostic process than even the most advanced algorithm can. Specifically, radiologists look at patient medical images in the broader context of examining and treating the whole patient, and thus provide a level of insight well beyond the framework of AI. Until that day arrives, we encourage diagnostic imaging providers to be diligent in conducting an AI cost–benefit analysis, as you would clearly do with any new technology.
Learn more:
AI Features in Radiology You Can Adopt Today
AI-Driven Smart Noise Cancellation
References:
- Auntminnie.com; AI needs robust clinical evaluation in healthcare.
- Advisory Board; “The Artificial Intelligence Ecosystem”, 2018
Jim Sehnert , PhD, is the Director of Advanced Development, Imaging Systems at Carestream Health.
Charlie Hicks is the General Manager for Global X-ray Solutions at Carestream Health.
