4 ways imaging informatics is central to AI in radiology

Artificial intelligence is in prime position to improve the field of radiology, impacting everything from image interpretation to clinical workflow. In order to be successful, radiology must lean on imaging informatics experts for help, argued authors of a new perspective published in Academic Radiology.

“An imaging informaticist is a unique individual who sits at the intersection of clinical radiology, data science and information technology,” Tessa S. Cook, MD, PhD, with the University of Pennsylvania, wrote in the review. “With the ability to understand each of the different domains and translate between the experts in these domains, imaging informaticists are now essential players in the development, evaluation and deployment of AI in the clinical environment.”

Below are four ways imaging informaticists can help successfully integrate AI into radiology:

1. Properly preparing data

Developing AI involves curating, labeling and augmenting large amount of data, a time-consuming and resource-intensive process. Radiologists are often the only experts that can do so, but they are already overextended—that’s where an informaticist comes in.

“Data curation, processing, labeling, and deidentification are fundamental problems tackled in imaging informatics,” Cook explained.

Projects such as the Digital Imaging Adoption Model—a collaboration between three large informatics societies—can help guide imaging facilities on how to identify and process data for AI creation. Informatics experts can also help deidentify radiology data contained in imaging reports to securely train algorithms on quality, real-world images.

2. Technical knowledge of the AI landscape

More often than not, those who are creating AI aren’t familiar with imaging informatics nor the challenges of dealing with DICOM data, let alone clinical radiology. Radiologists can work with informaticists to gain a more well-rounded picture of an AI project.

“Domain expertise is critical to the success and adoption of AI tools, both within and outside medicine,” Cook wrote. “Within radiology in particular, data scientists must learn both the clinical context for the problem being addressed as well as the technical aspects of the data, how it is created and stored, how to consume it, and what it represents.”

3. Evaluating AI

A majority of AI algorithms are not trained on external data, which can severely limit their clinical use. Imaging informaticists can help evaluate AI tools in the clinical environment—a task that is often too technical and time-consuming for radiologists. 

“This (clinical testing) requires careful thought and planning as to how to temporarily deploy the solution within the clinical workflow, most likely refraining from archiving results in PACS, but still giving radiologists a bona fide experience of using the tool on actual patient cases,” according to Cooke.

4. Integrating AI into clinical workflow

Moving AI from the research and development phase to a radiology department is no easy task, and requires “seamless” integration into the workflow without hampering the PACS or EMR.

Imaging informaticists are well-versed in the challenges and limits of deploying a new tool, patient privacy requirements and the importance of quality control checks.

Overall, Cook emphasized the need for radiologists to work hand-in-hand with informaticists to make sure AI doesn’t become another thorn in the side of imaging experts and their patients.

“It is our responsibility as radiologists and imaging informaticists to ensure that this new technology functions as expected, does not harm our patients, and improves the quality, efficiency, availability of and access to care for our patients,” she concluded.