A Quick RSNA 2023 Rundown
At Aidoc, RSNA is always a very busy time of year. For 2023, we prepared with a larger booth space with greater meeting capacity and, somehow, fielded several hundred more booth participants than in 2022. Having the opportunity to speak to this large swath of industry leaders helps inform how we continue to hone our view both in the near and long-term for the future of artificial intelligence in healthcare.
A handful of observations from this experience were immediately available to our team:
Bigger Was Better
It’s a good thing we expanded our booth presence for 2023. Every inch of our space was heavily leveraged. This included five different, separate meeting rooms, six different presentation stations, four couches, three different meeting tables, a greeting desk, and one extremely occupied barista. Kudos to all those providing coffee during this meeting. We are collectively indebted to you.
Larger Groups and Quality Questions
Many of the organizations who elected to meet with us brought entire teams into their scheduled discussion. With the addition of multiple stakeholders brought into each conversation, we saw a plurality and richness to the nature of the questions we were asked which I believe is a clear and strong indication of a maturing that is occurring in the market. These questions were not simply about how image interrogating AI works to find one particular pathology on one particular study. Rather, our discussions revolved around how to leverage AI generally for the sake of each patient. Or how to leverage multiple algorithms in the detection and triage of a particular disease. Further, our conversations often evolved into how Radiology continues to be the trusted supporting partner, the doctor’s doctor, for so many of their colleagues in other subspecialties. We believe quite strongly that, when leveraged properly, artificial intelligence tools can be implemented (our clients are testament to this) in a manner in which they assist the patient across the acute to subacute care continuum. Such was the nature of many of our client/prospect discussions.
Tomorrow’s Clients are Educated Buyers
Those considering the implementation of artificial intelligence tools in the near future are rapidly educating themselves through venues like RSNA on many of the critical factors which will determine the success or failure of their proposed projects. The new vernacular of this education are terms like dynamic orchestration, drift mitigation, clinical care coordination, data normalization, and, of course, aiOS™. The timeliness of this curricula cannot be over emphasized, and, while Aidoc is happy to play what role we can, we are just a party of one. As the saying goes, it takes two to tango. At this stage in the industry, we are happy to include many of the industry leading, luminary organizations in the world as our dance partners. It is exciting to learn, iterate, create and innovate with so many dedicated professionals focused on how to advance healthcare in a new and rewarding direction which showed up in full force in Chicago this year!
So What’s Next? Predictions for 2024
In the attempt to answer this very important question, I have three observations:
1. It’s time to go multimodal.
Great image interrogating artificial intelligence is really exciting. But do you know what is even more exciting? Automatically pairing those findings with their corresponding clinical information from the EMR. As everyone knows, clinicians today are completely overburdened. Part of this overburdening comes from how much work is involved in synthesizing vast quantities of information for every single patient. This is where we think artificial intelligence should play a critical role in the distillation and synthesis of known data such that our clinicians don’t have to go hunting for it each and every time a specific pathology presents itself. This type of rote labor will become less necessary over time as we implement more use cases which corroborate and match clinical information for each patient.
Therefore, in the extreme near future it will become crystal clear that in order to accomplish the automation needed for many of the labor intensive tasks physicians have today requires a platform which can bridge across the Imaging and Clinical archives. For example, if a small, asymptomatic, incidental aneurysm is found on a patient, the imaging finding alone may not be enough information to properly risk assess this particular finding. If the same patient has a history of smoking, high blood pressure, and a familial history with vascular disease, his/her risk profile can change entirely. Thus, the need to combine acute findings in their correct historical context is necessary in order to alleviate the burden placed on so many of our physicians today. Doing so will simultaneously raise the quality and the richness of the content of radiology reporting.
3. Therefore, governance takes center stage
As larger groups come together with better questions for vendors such as Aidoc, their ability to define their own destiny for the types of platforms they’ll permit into their institutions can begin to take shape. Many existing clients and perspectives inform us today that they have and/or are in the process of forming AI Governance Committees. While this often presents a vendor with additional steps in the process of becoming a partner, this critical decision making infrastructure is coming at just the right inflection point in the market in which technology is maturing along with the regulatory framework which will govern it.
The Regulators Have Arrived
While we have all been busily working to promote, implement and create new products and services, Federal regulators have had their eye on this space the whole time.
In the Spring of this year the ONC introduced HTI-1 (Health Data, Technology, and Interoperability final rule), which seeks to provide a regulatory framework for the technologies of tomorrow where there is…checking notes…no current regulatory framework.
Aidoc expects to receive and update on these guidelines intended to be leveraged for products, “…used to produce an output or outputs related to, but not limited to, prediction, classification, recommendation, evaluation, or analysis.” This includes technologies like vision enabled ChatGPT, Google’s Gemini or even the foundation model Aidoc has just invested $30 million to create – the first of its kind for healthcare. So while all the creative ideas and innovative functions related to the use of LLMs in our space are fantastic, they need a framework in which to function in production in our heavily regulated industry. Aidoc has led the way in the production of novel goods and services (17 FDA approved solutions so far) within this space and very much looks forward to continuing to provide leadership to the best of our ability.
