Radiology Coding Alert

Providers order and use imaging scans to view inside the patient’s body to diagnose unseen conditions and screen for diseases, but human physicians can miss abnormalities while reviewing the imaging scans. This is where artificial intelligence (AI) can assist radiologists in identifying conditions that went previously unnoticed.

Learn how radiology is locating hard-to-spot abnormalities with AI’s help.

Use AI With Imaging Scans

AI’s potential to assist physicians in evaluating X-ray, computed tomography (CT), and magnetic resonance imaging (MRI) scans and detecting conditions earlier than usual is making providers sit up and take notice. The American College of Radiology (ACR) published a Data Science Institute Artificial Intelligence Survey in 2021 (www.sciencedirect.com/science/article/abs/pii/S1546144021002933). The survey showed significant growth of AI adoption in radiology from 0 to 30 percent over a five-year period. Also, approximately 20 percent of respondents planned to purchase AI tools within the next one to five years.

Two years after the ACR published its survey findings, radiology is outpacing other specialties when it comes to exploring AI’s capabilities. According to the U.S. Food and Drug Administration (FDA), there are more than 690 market-cleared AI medical algorithms available in the U.S., as of this publication (www.fda.gov/medical-devices/software-medical-device-samd/artificial-intelligence-and-machine-learning-aiml-enabled-medical-devices). Of those 690+ algorithms, 531 are related to radiology alone — approximately 76 percent — while cardiology accounts for 71 algorithms, and 20 algorithms are related to neurology.

Examine the AI and Radiology Intersection With Research Studies

Similar to AI’s use in the revenue cycle, radiology is a specialty where AI can serve as a helpful tool for providers. Adding AI to the interpretation portion of an imaging test can assist radiologists in identifying abnormalities, precisely categorizing the anomalies, and pinpointing the issues quickly — below are two examples.

On March 7, 2023, Copenhagen, Denmark researchers published their findings on the effectiveness of autonomous AI in examining chest X-rays (CXR) (https://pubs.rsna.org/doi/10.1148/radiol.222268). They used consecutive posteroanterior (PA) CXR from 1,529 adult patients obtained in January 2020. The patients were from emergency departments, inpatient, and outpatient facilities. A commercially available AI tool analyzed the X-rays and classified the images as “high-confidence normal” or “not high-confidence normal,” also known as normal and abnormal, respectively. The AI tool identified the abnormal CXR correctly 99.1 percent of the time.

On May 22, 2023, researchers at New York University (NYU) Grossman School of Medicine and NYU Langone Health published their findings showing that AI successfully identified brain structure changes resulting from repeated head injuries (https://journals.sagepub.com/doi/10.1177/19714009231177396). Using hundreds of MRI brain scans of male college athletes captured between 2016 and 2018, the AI system and machine learning (ML) accurately identified small structural changes in the participants’ brains. The brain scans with the structural changes came from athletes who played contact sports, such as football, but had not received a concussion diagnosis.

“Our results highlight the power of artificial intelligence to help us see things that we could not see before, particularly ‘invisible injuries’ that do not show up on conventional MRI scans,” said Junbo Chen, MS, PhD candidate at NYU Tandon School of Engineering in a press release. “This method may provide an important diagnostic tool not only for concussion, but also for detecting the damage that stems from subtler and more frequent head impacts,” Chen added.

The NYU researchers are planning to conduct another study with female athletes.

Could Reimbursement News Encourage Further AI Adoption?

While researchers have found that AI shows immense potential in evaluating imaging scans, the technology is not without its challenges. Healthcare providers are reticent to adopt AI for their practices for several reasons, but one of the biggest hurdles to adoption is reimbursement. Practitioners want to know how much they’ll be paid for the services they provide, and while AI’s abilities are promising, there are only a few examples of AI generating reimbursement to assure the cost of the investment. But change could be right around the corner.

On June 30, 2023, the American Medical Association (AMA) announced the addition of two CPT^® Category III codes, which took effect Jan. 1, 2024. The two codes use AI-related brain MRI quantification software to help clinicians diagnose, monitor, and evaluate treatment options for different brain disorders.

The new CPT^® Category III codes are:

• 0865T (Quantitative magnetic resonance image (MRI) analysis of the brain with comparison to prior magnetic resonance (MR) study(ies), including lesion identification, characterization, and quantification, with brain volume(s) quantification and/or severity score, when performed, data preparation and transmission, interpretation and report, obtained without diagnostic MRI examination of the brain during the same session)
• +0866T (… obtained with diagnostic MRI examination of the brain (List separately in addition to code for primary procedure))

Healthcare providers can use the quantitative MRI analysis software with patients with Alzheimer’s Disease and Dementia, epilepsy, traumatic brain injuries, and strokes.

“To date, only a handful of companies, in the field of heart disease, lung cancer, and liver disease, have successfully obtained a CPT^® III code in radiology AI. As this is the first CPT^® code for quantitative brain MRI analysis, we are thrilled about the impact this code will have on the care and management of patients with brain disorders,” said Dirk Smeets, chief technology officer at icometrix in a press release (https://icometrix.com/news/ama-issues-cpt-iii-code-for-icometrix-ai-related-brain-mri-quantification-software).

Note: Medicare and other payers may not reimburse you for Category III codes. However, reporting Category III codes could contribute to creating future Category I codes that will lead to payment. Category III codes are temporary codes and consist of four numbers and end with a “T” to signify their temporary status. According to the AMA, Category III codes are “intended to be used for data collection to substantiate widespread usage or to provide documentation for the Food and Drug Administration (FDA) approval process” (www.ama-assn.org/practice-management/cpt/category-iii-codes).