In radiology, AI has moved from a buzzword to a valuable tool: RSNA 2022

By: Richard Hausmann, CEO; Lennart Thurfjell, COO; Greg Kingston, CMO

In the AI Showcase at the 2022 Radiological Society of North America (RSNA)’s scientific assembly and annual meeting, we joined 118 other exhibitors representing companies providing artificial intelligence (AI) products and services to support the entire imaging workflow, from data preparation to reporting and everything in between.
Registration for this year’s conference reached nearly 35,000 people, more than a 60% increase over last year. Although not quite to the nearly 50,000 registrations recorded pre-COVID, it was refreshing to see the in-person traffic and what seemed to be nearly equal representation from US and European colleagues. We enjoyed the conversations we had in our booth and beyond. From these, we identified a few common themes, described below.

AI is increasingly accepted as a valuable tool in imaging.

Radiologists worldwide are seeing the value of AI for improving the quality and accuracy of medical imaging, as well as streamlining aspects of their workflow. Although that acceptance is still in the earlier stages for our European colleagues than our US colleagues, as we heard at the European Congress of Radiology (ECR) 2022 and again at RSNA 2022, we continued to have meaningful conversations with all attendees about the practical value of AI, how it would fit into existing workflows, and its impact on patient care.
In fact, in our discussions with radiologists, we’ve noticed a shift from apprehension about adopting AI to determining how it could best address specific challenges experienced in daily work, such as the arduous task of counting lesions and calculating lesion volumes, as well as their change over time, in multiple sclerosis (MS). As we learned earlier this year at Americas Committee for Treatment and Research in Multiple Sclerosis (ACTRIMS) 2022, AI applications are playing a crucial role in the earlier diagnosis and ongoing monitoring of MS by identifying radiographic characteristics that are difficult for a human to detect and quantify.

Partnerships will likely be key to minimize disruption to workflows.

With similar efficiencies being realized across other neurological diseases as well as oncology, emergency, and other therapeutic areas, the challenge remains how to seamlessly integrate AI into existing workflows. The benefit of saving time to evaluate images is for naught if more time is lost switching between platforms or manually transferring AI results into standardized reports. End users are interested in solutions that support their true end-to-end workflow, from reading images to populating the results into commonly used structured reports as well as patient records.
One trend we observed last year at RSNA 2021 was the growing use by larger vendors of platforms that combine numerous proprietary and third-party algorithms and applications in a single location. The presence of this vendor-agnostic platform approach was also evident among exhibitors in the AI Showcase and beyond at RSNA 2022.
For radiologists and technicians, an advantage of this approach is the central location to use AI tools from different providers in an integrated fashion. The platform vendors benefit from the ability to integrate their own core applications with innovative or specialized AI applications from other companies, while eliminating the investment of developing and maintaining all the applications themselves. For companies developing AI solutions, offering their product via a platform could potentially help:
This can be particularly relevant for end users who would otherwise be reluctant to switch vendors; it can occur within the platform with minimal disruption to their work.
At the same time, we observed that some larger companies placed less focus on their AI platforms at this year’s conference than in previous years, potentially because of the complexity of managing multiple applications and vendors. As these companies retrench to find a better solution, an alternative approach that appears to be gaining popularity is offering a suite of tools, such as a neurology, oncology, or cardiovascular suite, that integrates the algorithms and applications relevant to that therapeutic area. This reduces the overall scale of the platform and responds to users’ recent demands for a more focused solution.

There is growing interest in accessing deeper, broader patient data.

We heard it last year and again this year: data is everything. Radiologists and their referring clinicians are interested in accessing the broad range of data that is truly needed for better understanding of disease mechanisms and risk factors, earlier diagnoses, monitoring disease progression and treatment effects, and predicting disease progression.
In neuroradiology and neurology, being able to incorporate data such as imaging and other biomarkers, genetic information, clinical information, and patient demographic data is becoming increasingly important for early diagnosis and disease monitoring of MS and dementias and for clinical treatment of dementias as new disease-modifying drugs (DMDs) for Alzheimer’s disease are developed. For the latter, this information is especially crucial for identifying eligible patients as well as monitoring treatment effects and side effects.
One challenge with increasing the amount of data used is the ability to process it efficiently and quickly so the results are immediately actionable for a patient, which is nearly impossible for a human to do. This is the perfect application of AI — identify patterns in a patient’s data and compare those with reference data sets to determine the most likely diagnosis and best treatment course. We designed our cDSI™ application with this in mind — going beyond detection and volume measurement using AI to assess key data for patients with neurodegenerative diseases. The results provide support for early disease detection, differential diagnosis, the most likely disease progression, and long-term monitoring.

There is always scope for improvement.

Insights that we gain at conferences usually spark ideas for future enhancements to our applications — whether it’s further developing our AI or incorporating changes that make it easier for radiologists or clinicians to do their job.
Being able to automatically detect secondary findings on imaging was something we found interesting. Although not necessarily something we might develop ourselves, we could immediately see the relevance of this for patients with neurodegenerative disorders, who can often have comorbid conditions. In relation to DMDs, this type of processing could be beneficial for detection of side effects on MRI, proactively alerting the radiologist of at-risk patients.
Over the past year, we’ve focused on improving the clinical relevance of our reports after many conversations with clinicians, colleagues at conferences, and our partnering institutions. Based on feedback from booth visitors at the last couple of conferences, we’re feeling pretty confident that we’ve achieved our goal! Our unique Dementia Differential Analysis report is driven by our cDSI’s proprietary machine learning method to offer true differential diagnostic support for dementias using MRI data. Our new clinically focused reports for dementia, MS, traumatic brain injury, and epilepsy present clinically relevant imaging information for each of the four conditions in the form of annotated images, percentile graphs, and color coding of values that fall outside of the normal reference range.

PET imaging has a role, even at a radiology conference!

We were pleasantly surprised by the amount of interest in our new cPET™ application (not yet approved for clinical use) at a radiology conference. To us, it demonstrated how radiologists are interested in a comprehensive set of information to support their referring clinicians in providing better patient care. cPET provides fully automated quantification of FDG and amyloid PET tracers, which complement other patient information for improved diagnosis and management of dementia and other neurodegenerative diseases — especially around patient eligibility and side effect detection with DMDs for Alzheimer’s disease.

Enhancing imaging enhances patients’ lives.

According to RSNA President Bruce G. Haffty, MD, “To the patient, imaging can remove uncertainty, decrease anxiety and give hope. Through the panoramic lens of the patient, imaging is a powerful and meaningful source of knowledge and comfort that we can all relate to.”
As an industry, we’ve made huge strides in the ability to use technology to provide more relevant, higher-quality imaging results, and we have many more opportunities to extract greater value from imaging through AI in the future. As we settle back into our daily routines, we’re looking forward to tackling the next set of challenges with our applications.

Interested in learning more about our applications, or have feedback of your own? Contact our team.