The Role of Artificial Intelligence (AI) in Transforming Medical Imaging
By Tanya Dodge, Director of Imaging Services, HCA Midwest Health
For one moment, indulge me, close your eyes, and think what it would be like fifty years ago to wake up one morning normal, having a stroke, and your whole life changes. You suddenly can’t walk because you are paralyzed on one side; you can’t talk except to mumble a couple of the same words over and over, and no one knows what you want, let alone have a way to help you find a way to communicate. You are stuck inside yourself, and your only option is to learn to live like this new reality. Now, fast forward fifty years later, and if you have a stroke today, we rush you to the closest certified stroke center and launch your care in a CT scan with a neurologist, most likely online; AI algorithms read and analyze the exam with a report in less than two minutes, then administer medication and, in most cases, you get to walk out like the old you. Admittedly, this is a very simplified version of how AI advancement helps change the course of our patient’s lives and lessen the burden on staffing. Still, it is the new way of medical care for thousands, moving from slow, inefficient care to fast, quality results without adding a workforce we all don’t have.
AI advancements are helping facilities, even small critical access hospitals, combat staffing shortages, lack of expertise, volume overload, and the ability to solve some medical inequality. This is a bold statement, but we must find ways to staff that don’t involve more humans. They are not simply there to hire, the pandemic being the primary reason for thousands leaving the field. AI is simply an algorithm that runs in the background and can be utilized with most HIM and PACS systems to help improve throughput, sort exams by severity, and prioritize the most critical patients for the radiologist to read first. In some incidents, it can flag the slice the reader wants to verify and pop it to the top of the radiologist list. Seminars tend to focus on advancements in healthcare, and the speakers in imaging are no different. They have all addressed AI, how we are using it to combat the loss of staff, and how to handle the growing volume of exams physicians are ordering to diagnose patients.
AI can be used in almost any imaging modality. Still, most AI is heavily concentrated in the CT modality since CT is the most utilized imaging department, especially in an emergent setting. Stroke AI care has been the grandfather of AI, and we have long used it to help deliver best-practice stroke care to gain certification. But now, AI is advancing, and we have added deep AI. The two types of AI are allowing us to reach more patients and learn from what we know and what we will learn from its vast data collection ability. No human, even a radiologist, can collect and calculate data at this speed compared to AI. No radiologist will be replaced, but it will allow the staffing shortage of radiologists to focus on quality reading and less on quantity reading. Studies out there prove that when radiologists go faster, it leads to mistakes and missed diagnoses.
AI can pay for itself if it generates enough incidental findings that would have otherwise been missed.
AI algorithms are used to look for specific pathologies. AI is often used emergently for high acuity exams like pulmonary embolism (PE), aortic dissection, cervical fracture, and multiple others to choose from. Humans can’t read hundreds of exams fast enough to get critical results to the ordering provider within minutes, but AI can. The list of what you can purchase is vast, and cost is often based on volume or clicks. I have even worked with overseas companies on some AI for fractures in diagnostic bone work since those exams can be left to last during patient surge times, and the radiologist’s time is focused on reading imaging of higher levels of care. These routine diagnostic exams are often not seen as high acuity but still require the radiologist’s attention.
Patients demand faster, quicker, and more accurate care, and having 27% faster care is promising. It is becoming increasingly more complex to deliver high-quality care at affordable prices and improve population health without all the experienced staff we had before COVID-19. Healthcare staffing and care is a struggle, whether you are a small critical access facility or a large corporate group of facilities. Facilities are all struggling and must look at creative ways to focus on the Triple AIM of healthcare. Several experts agree that there isn’t any way to do that without employing AI and technologies to speed up human work processes. Suppose healthcare workers are to sort through the hundreds of daily exams to figure out who has the critical result, fractures, or what patients need to be rushed to the operating room (OR) or emergently admitted due to a critical diagnosis like a PE. If fifty patients need a PE chest CT read, and they have the positive scans that happen to be #39 and #48, then they wait until the radiologist gets to that particular exam, but not with AI; they go to the top in minutes, not hours. To determine the next steps, a facility will have to determine its strategic plan, weaknesses, and where they can deploy AI to best improve throughput for their patients.
A facility must determine which AI company best aligns with its goals and how it will pay for the technology. A consideration to remember is that AI can also help you find more work by finding incidental findings to follow up on in an outpatient setting. Thus, AI creates more work but more revenue to pay for itself. Depending on your facility’s goals, this can be a primary or secondary goal of your AI journey. AI can pay for itself if it generates enough incidental findings that would have otherwise been missed.
