From enhanced image quality and workflow efficiencies to an improved patient experience and potential synergies wih enterprise cloud services, artificial intelligence continues to redefine possibilities in radiology.
There are numerous technologies and technical terms related to artificial intelligence (AI) that can distract some people from the rationale behind pursuing AI solutions. At Carestream, we are crystal clear on why we are applying AI to our medical imaging solutions. For us and the customers we serve in radiology, the “why” behind AI is improving clinical outcomes and patient care, and creating more space and time for human interaction.
Often, imaging is the first step to making an informed diagnosis. Medical images are quantifiable. They can either prove or disprove that something is wrong, and both results are equally valuable. The amount of quantifiable information within a digital X-ray is tied directly to the quality of the image.
We are intent on providing as much detail as possible in a digital X-ray – at the lowest possible dose – so radiologists have the information they need to make a confident diagnosis. Our AI solutions play an important role in delivering on this goal. For example, our Eclipse Imaging Intelligence capabilities deliver superb image quality and unrivaled diagnostic confidence with AI, proprietary algorithms and advanced image processing capabilities.
Additionally, our Bone Suppression software leverages AI to suppress the appearance of bone to enhance the visualization of soft tissue while requiring no additional exposure to the patient. “Removing” the bone from the image gives the physician a clearer image of the area of concern, and helps inform his or her assessment of the pathology.
In medical imaging, the goal of capturing the most information possible in an image must be balanced with the need to limit excessive radiation dose. Our AI-powered Smart Noise Cancellation (SNC) software advances this objective by allowing a lower radiation dose without a loss in image quality in comparison to standard image processing.
We are also applying AI to improve the process for capturing precise, quality images. Mistakes can be made in the image capture process for several reasons, including incorrect positioning of the patient and/or the system. That’s why Carestream is applying AI to automate these steps in the image acquisition process.
(Editor’s note: For related content, see “Assessing the Value Proposition of AI in Radiology,” “AI in Radiology: Top Five Articles of 2022” and “Maximizing Cloud-Based Capabilities in Radiology.”)
Our AI-based Smart DR Workflow enables a more precise capture of the anatomy needed to make a proper diagnosis. In addition to helping radiology teams capture the best and most quantifiable image possible, the automation also makes the process more efficient. This gets diagnostic images into the hands of the radiologists and physicians as quickly as possible so they can begin a course of treatment for their patients.
Getting the image captured properly the first time also reduces the need for X-ray retakes, thus limiting unnecessary radiation exposure. Indeed, AI software can play a considerable role in improving clinical outcomes by delivering superb image quality at the lowest possible dose, and enhancing the precision of the image capture process.
How AI Can Improve the Image Capture Process for Patients and Radiologic Technologists
Carestream’s AI software is also designed to improve the patient experience by creating more space and time for human interaction.
The image capture process is an interesting juxtaposition of leading-edge technology and a very interactive human process. I noted earlier that the automation enabled by our AI-based Smart DR Workflow allows a radiologic technologist to perform an exam in less time. Spending a minute or two less on a cold hard imaging table might not matter much to a patient with a fractured arm. However, having to hold a difficult position for a minute or two less can be significant to a patient with a painful injury or someone who has anxiety or confusion about the procedure.
One should also consider the impact for the radiologic technologist. The AI workflow features give technologists added confidence that they are capturing the best image possible that will help advance the patient’s care. Bringing AI (automation) to the process also frees up radiologic technologists to focus on one of the meaningful parts of their job: interacting with the patient. Perhaps they are freed up to hold an elderly patient’s hand for a minute longer. This small but meaningful human interaction also can improve the patient experience.
Maximizing the Capabilities of Enterprise Cloud Services to Accelerate Research and Testing of AI Models
Our “why” for applying AI to radiology is to help improve clinical outcomes and patient care. Now let’s delve into the “how.”
In the simplest of terms, machine learning is a branch of AI that focuses on the use of data and algorithms to imitate the way that humans learn, gradually improving their accuracy. The more data that is ingested, the more the algorithm learns. The “data” in this case may be X-ray images. In order to develop the smartest and most reliable AI applications possible, we train our machine learning algorithms on thousands of de-identified diagnostic images provided through agreements with our customers.
As you can imagine, this requires massive computing power and time. That is why we entered into a strategic partnership with Hewlett Packard Enterprise (HPE) GreenLake to leverage its enterprise-grade cloud service for machine learning (ML Ops platform).
The platform enables an accelerated workflow for research, including testing AI models on clinical data, getting faster feedback, and deploying better solutions. For example we had a significant reduction in time to execute training runs for our AI-driven Smart Noise Cancellation solution, going from 60 hours to 16 hours.
What is the connection between the HPE platform (the “how”) and our “why?” It accelerates our delivery of AI-powered solutions that can help our health-care customers to impact patients’ care. Using the HPE platform also frees up our imaging scientists to focus more time on creating solutions for our customers while spending less energy on building the “plumbing” needed for AI solutions creation.
Our use of the ML-Ops platform is in its formative stage at the moment. As we expand the use of AI techniques to meet the growing clinical and operational needs of radiologists, we will continue to bring in more researchers, engineers, and clinicians to create these solutions. We believe we are just scratching the surface of how AI and machine learning can help improve patient outcomes and patient care, and that is our ultimate “why.”
Editor’s note: This article was adapted with permission from its original publication on Carestream’s Everything Rad blog at https://www.carestream.com/blog/2022/06/07/the-why-behind-applying-ai-in-radiology/