By Matt Skoufalos
Just as the advancement of medical imaging technologies has evolved nearly every aspect of healthcare delivery in the modern era what it previously could achieve, so too has the refinement of artificial intelligence (AI) and AI-powered solutions compounded those gains, paving the way for new growth in the systems that allow healthcare practitioners access to the inner workings of the body.
AI technologies like natural-language search, machine learning, and vast data-analysis tools have found ready utility in the world of medical imaging, facilitating the heavy computing needs of image-capture, processing and data storage systems that underpin the operations of medical imaging systems. The earliest promised returns on the synthesis of these technologies have been realized in improved radiologist workflow, patient scheduling and image refinement processes. Discovering applications through which they may be leveraged in the future – namely, in improvements that can play a more direct role in the patient experience – involves taking a closer look at how AI-powered processes presently are deployed, and in what ways they may be.
Jason Polzin, general manager for MR applications platform and research technologies at GE HealthCare, agreed that much of the advancement of AI-powered technologies in the medical imaging space has focused on the experiences of the radiologists and technologists that most frequently interact with imaging devices themselves.
Polzin pointed out, however, that as much as AI computing advancements that improve image quality, reduce length of scan times and automate patient positioning supports the work of imaging professionals, they also can improve the patient experience during an exam by streamlining those various interactions into a smoother process.
Moreover, he said, “where there’s overlap” in the realized gains of AI technologies for a variety of stakeholders, “that’s best.”
“The best use of AI is where it hits administrators, clinicians and patients,” Polzin said. “A lot of what these AI technologies do is reduce recalls and re-scans. When they have to come back a couple days later, that’s very disruptive.”
“Shortening the amount of time also makes the exam better for the patient because the less time they’re on the table, the better an experience it is for them,” he said. “It’s less time being anxious and having to hold still.”
Likewise, AI-powered intelligent protocoling that gathers personalized patient data from electronic medical records (EMR) and prior examinations helps inform automated processes onboard the imaging devices themselves about what studies may be performed based on prior experiences. Whether those solutions are integrated within the platforms that vendors create themselves, or added as an after-market enhancement, the value they add in the clinical domain all distills into a more seamless patient experience.
“When we consider workflow, the more integrated the solution, the easier it is to adopt, and for the scan operator to use,” Polzin said. “It’s critical that the on-device, uses [of AI technologies] are available. This includes developing systems that are compatible with third-party solutions being leveraged for their deep learning and AI expertise in the clinical domain.”
“We want this to be as seamless a workflow as possible,” he said. “We spend as much time developing these AI technologies to make the clinician’s job easier, as we do working to ensure seamless integration for the radiologist, the technologist and the patient.”
Polzin’s colleague Erdogan Cesmeli, GE HealthCare chief strategy, marketing, and commercial officer of molecular imaging and computed tomography, said it’s easiest to contemplate the myriad ways in which AI-enhanced technologies support the patient experience by regarding the patient journey from a holistic perspective. AI processes can accompany the patient from ordering an imaging study to scheduling it, undergoing the exam, communicating and analyzing the results afterwards, and maximizing the impact of those findings.
“[The patient] starts out seeing their generalist, who may be referring them to their cardiologist or radiologist, and then get scheduled for CT or MR,” Cesmeli said. “During that process, we get their demographics, some of which are used with our images.”
AI can aggregate detailed patient data like height, weight, age and gender to support optimization of imaging protocols applied at the device level during an imaging exam. Technologists and physicians benefit from access to the menu of those options, and ultimately must decide which is best to use for which patient; however, access to that combined knowledge base can help to optimize the image reconstruction, Cesmeli said.
“More and more, with AI technologies, patients come to the room, they are positioned, they are scanned, and before they probably dress up, the images are analyzed, and sent to PACS,” he said. “Radiologists are then reading them and sending a report to the referring physician, who shares the results with the patient.”
“We are more [concerned] about [what happens] when the patient shows up to the room where our equipment is, and [how] the data goes to the reader or physician preparing the report [afterwards],” Cesmeli said.
“AI is a little bit like art when you are doing the reconstruction,” he said.
Optimizing patient data for imaging protocols can help expedite the process of undergoing an imaging study; optimizing the reconstructive analysis that helps radiologists and referring physicians process the results of these studies can have another spillover effect on workflow; namely, freeing up some of the time spent analyzing those reports.
“Deep-learning-based [image] reconstruction has really taken the industry by storm,” Polzin said. “The images are just easier to interpret and easier to look at, [so] a radiologist doesn’t have to spend as much time interpreting the images. The other thing is that some of these clinical decision support tools help clinicians to interpret the exams more quickly, consistently and confidently.”
When AI-powered image reconstruction, analysis and transmission to PACS processes are automated through technologies that are onboard an imaging system, they can eliminate hours of inefficiencies that free up healthcare professionals for other uses of their time, and potentially help alleviate burnout, Cesmeli said.
“The whole point is about giving back time to the reader, the physician, the radiologist so they are less prone to make mistakes, and patients can get results back faster,” he said. “Remove the burden on the reading physicians so they can get to the point with minimized risks and can provide the accurate diagnosis.”
Beyond their diagnostic imaging applications, AI-powered processes can also support imaging-guided patient treatments, like theranostics and radiation oncology. Both disciplines involve highly advanced applications of medical-imaging-guided processes that can benefit from AI computing solutions to refine and advance the effectiveness of the treatment processes.
In theranostics applications, PET-CT (Positron-Emission Tomography-Computed Tomography) provides detailed anatomical information that illustrates how tissue functions at a molecular level, the better to destroy cancerous material with a targeted radiation dose. Similarly, when delivering radiation oncology care, AI-powered processes can help automate treatment planning, support clinical decision-making, and deliver personalized care specific to any patient’s individual anatomy.
“If [physicians] see that there’s no response [to the treatment], they stop trying that treatment,” Cesmeli said. “There’s optimization from the patient point of view. We are in collaboration with the pharmaceutical companies, using AI to personalize some of those treatments even better.”
On the other hand, some AI-driven advancements are evolving so rapidly that manufacturers like GE HealthCare recognize the need for specialized, third-party integrations. After acquiring AI developer MIM Software in 2024, GE HealthCare was able to integrate its own cardiac imaging systems with advanced PET and SPECT (Single Photon Emission Computed Tomography) post-processing and analysis tools from MIM.
Cesmeli described strategic acquisitions like these as “not completely outsourcing” the AI functionality that underpins these processes but allowing GE HealthCare to be “the face and quality check of a major vendor in this space.”
“Partnership is a necessity, especially for those narrow-but-focused companies,” he said. “[Our customers would like] for us to be able to [provide] the platform so that we can easily include these [functions]. Instead of trying 10 different packages, they would like to have one they can trust, and they can do the risk assessment.”
Ultrasound imaging is another area in which the application of AI-powered processes can help to evolve and refine a mature, ubiquitous technology to improve patient wait times, limit repeat scanning, support image quality, and even help physicians to recapture revenues that may previously have been lost due to oversights in patient documentation.
As POCUS (Point Of Care UltraSound) continues to expand its reach in a space that is also facing a continued staffing shortfall, AI computing can help to automate certain aspects of a modality that relies greatly upon the manual skills of its users. It’s a process that Cesmeli said resulted from an outgrowth of manufacturers finding ways to automate cross-disciplinary functions among MRI and CT technologists.
“One of the things that we are doing is finding out how to make our user interfaces more intuitive, so that an MR tech can also do CT scans, and vice versa,” he said. “We think AI is going to be useful in that area versus the discrete layout.”
“Ultrasound is real-time imaging, so in that case, guided by the AI, [a sonographer finds] the point of interest, or doesn’t,” Cesmeli continued. “I think it’s very clever. It’s less so that we have an interventional mode for performing a biopsy, and more that we have another device that gives the same feedback loop whether to do the biopsy or not.”
Other emergent research focuses on the idea that AI-powered processes could help automate back-end reporting and data aggregation that improves the patient experience. A July 2025 study published in Scientific Reports describes how a research team, led by Oluwatosin Ogundare from the Department of Information and Decision Sciences at California State University-San Bernardino, developed a theoretical “AI Affinity Score” model to predict how AI integration can “maximize their experience of care.”
“Typically, the ultimate goal of the customer satisfaction survey is profit maximization for the service provider, whereas an assessment of care aims to cater to an optimization of the psychological factors within the context of the healthcare service,” Ogundare wrote. “One might argue that a comprehensive study of the patient’s experience of care should assess how much of the human qualities of empathy, understanding, gentle touch, etc., is lost in the AI integrated healthcare.”
The study concludes that AI-based therapy can be most effective for patients who trust it as a segment of their care delivery, accounting for differences in age, education and other socioeconomic factors.
If more closely integrating AI with patient care leads to better health outcomes or improved delivery of care, it stands to reason that patient satisfaction could also improve correspondingly. However, Ogundare, et al., noted that those outcomes would be best driven by an understanding of patients’ confidence in them.
Another study from the JANUS Group of Barcelona, published in the English edition of Medicina Clinica in February 2025, talks about how AI could improve outpatient care by incorporating “socio-health information on each patient to better personalize their clinical care.” Researchers believe this could include tracking health indicators through wearable devices, self-administered questionnaires, and other data sources to engage the patient in shared decision-making.
That study concludes with the following statement:
“In this sense, it is considered that AI should evolve from its current, relatively ‘narrow’ applications in healthcare (performing specific tasks repeatedly at the back of the medical consultation) to a broader and more flexible perspective that allows the current healthcare system (based on primary and specialized care) to be transformed into one that also includes the patient as a key player in the implementation.”
“There’s a psychology to this,” Cesmeli said. “People have different preferences [for the use of AI processes in different environments. Attitudes reflect that] it’s OK if artists use AI for music, but not for a politician to prepare their speech.”
“We hear from customers that they don’t want us to fully automate everything,” he said. “They want to automate the boring stuff, but leave the fun stuff so they can spend more time with their patients, and in quality [areas].”
The bottom line in all of these processes, as Polzin said, is that whether AI is employed in the calculus or not, “physician trust is critical” to the patient experience.
