The key to making AI work in radiology is not technology or corporations or even the FDA, it’s having the right leadership in place at the practice level, and without that, AI will never properly make its way into routine clinical use.
The goal with AI is to make practices more efficient, more sustainable and better overall. But to do that, a lot of ground work needs to be done. Radiology practice management will need to overcome several challenges to bring AI to clinical practice. It will be hard, it will be time consuming and at times frustrating, but for those that do, the rewards will be significant. Here are a few suggestions.
- Deciding who will be on the AI committee. Every radiology group should have a committee of radiologists who are sufficiently versed in AI technology and available AI based clinical applications.
- Picking how/when to use AI in your practice. As we all know, each radiology practice is different. A teleradiology group might get significant value from one clinical application, while a hospital-based group may benefit from a completely different application (hence why having a knowledgeable committee is important). Your practice needs to set priorities based upon the needs of your patient population.
- Reimbursement. Conversations need to start now with the major payers in your area. If practices can work with payers early in the process, they can find ways to streamline payment for clinical applications that both radiology groups and payers agree to.
- Onboarding. Bringing on AI is similar in some ways to bringing on a new radiologist. It may only be "licensed and credentialed" for certain exams from certain sites from certain payers. If you do not have a good way to send these exams through an AI enabled workflow, you will need to solve that problem before purchasing any AI clinical applications.
- Quality. AI clinical applications, just like any radiologist in your practice, needs to be constantly monitored for quality. A modified version of the peer review process will allow the practice to audit a certain percentage of the results of your AI clinical applications. Baselines of quality should be established at your practice, before rolling out any AI clinical application for broad use by the practice.
- Partnership with hospitals. Just like with payers, partnerships need to be developed with your hospitals and other customers your practice reads for. It will be important to check in with them frequently as AI clinical apps are rolled out. Each practice needs to gather statistics showing, for example, how TATs of positive exams are improving.
In some ways, practice managers will be on-boarding an "AI Robot" to assist with your practice. If practices leverage some of their existing tools, consider all stakeholders, and surround AI with the right workflow, I expect AI will be a boon to those practices that choose to embrace AI and lead its logical deployment.