Imaging AI Funding Still Solid in 2022

Despite plenty of challenges, imaging AI startups appear to be on pace for another solid funding year, helped by a handful of huge raises and a diverse mix of early-to-mid stage rounds.

So far in 2022 we’ve covered 18 AI funding events that totaled $615M, putting imaging AI startups roughly on pace for 2021’s record-high funding levels ($815M based on Signify’s analysis). Those funding rounds revealed a number of interesting trends:

  • The Big Getting Bigger – $442M of this year’s funding (72% of total) came from just four later-stage rounds: Aidoc ($110M), Viz.ai ($100M), Cleerly ($192M), and Qure.ai ($40M), as VCs increasingly bet on AI’s biggest players. 
  • Rounding Up the Rest – The remaining 14 companies raised a combined $173M (28% of total), with an even mix of Seed/Pre-Seed (4 rounds, $10.5M), Series A (5, $74M), and Series B (5, $89M) rounds. 
  • VCs Heart Cardiovascular AI – Cardiovascular AI startups captured a disproportionate share of VC funding, as Cleerly ($192M) was joined by Elucid ($27M) and Us2.ai ($15M). Considering that Circle CVI was recently acquired for $213M and HeartFlow has raised over $577M, cardiac AI startups seem to have become imaging AI’s valuation leaders (at least alongside diversified and care coordination AI vendors).
  • No H2 Drop-Off (yet) – The funding breakdown between Q1 (6 rounds, $63.5M), Q2 (7, $289M), and Q3 (5, $263M) doesn’t suggest that we’re in the middle of a second-half slowdown… even though we probably are. 

The Takeaway

Despite widespread AI consolidation chatter in Q1 and the emergence of economic headwinds by Q2, imaging AI startups are on pace for yet another massive funding year. These numbers don’t reveal how many otherwise-solid AI startups are struggling to secure their next funding round, and they don’t guarantee that funding will also be strong in 2023, but they do suggest that 2022’s AI funding won’t be nearly as bleak as some naysayers warned.

Imaging IT’s Infrastructure Problem

A heated Twitter conversation revealed widespread discontent with imaging’s outdated and fragmented IT infrastructure, suggesting that it’s draining radiologist productivity and standing in the way of AI adoption.

This tweet by Memorial Sloan Kettering’s Anton Becker, MD, PhD got things started: “95% of radiology departments would do well to direct 100% of their AI efforts and budget towards upgrade and maintenance of PACS, RIS and dictation software for the next 5 years… Our field is plagued by legacy software.” 

And here’s what the ensuing replies and retweets revealed:

  • PACS Productivity – Nearly everyone agreed that their overall imaging IT setup was insufficient, with one rad estimating that a “supercharged PACS” would improve his productivity by 30%, and another noting that workflow customization would “at least double” her speed and accuracy.
  • Imaging IT Revolution – Some called upon the “legacy” PACS, RIS, and voice recognition vendors to make more “revolutionary changes,” rather than settling for tweaks to current setups. Others proposed government intervention.
  • IT Isn’t Flashy – One thing that might be holding some imaging IT overhauls back is “it’s not as flashy to boast” about high-quality infrastructure, and “the people who have authority to allocate resources are more motivated by flash than function.”
  • Holistic IT – Eventually the conversation led to several well received proposals that we “eliminate the idea of PACS as a category and start thinking more holistically about radiology IT.” In other words, this might be more of a “fragmentation problem” than a PACS/RIS/voice functionality problem (or an AI budget problem). 

The Takeaway

Even if RadTwitter tends to skew towards academic radiologists and often focuses on what’s going wrong, this conversation indicates widespread dissatisfaction with current imaging IT setups, and suggests that radiologist productivity (and perhaps accuracy and burnout) would improve significantly if imaging IT worked as they’d like it to work. 

It’s debatable whether this imaging IT problem is actually due to an unnecessary focus on AI (very little of the conversation actually focused on AI), but it does seem reasonable that rad teams with solid infrastructure would be more likely to embrace AI.

Technologists in the Spotlight

Radiographers and technologists were at the center of this week’s radiology news cycle, as three unrelated pieces highlighted the crucial role radtechs play, the significant challenges they face, and the actions required to help them succeed.

CXR Call to Action – After finding that nearly half of their portable chest X-ray images were “problematic,” a team of Stony Brook physicians issued a “call to action” to better support radiology technologists. Analysis of 500 portable CXRs found 231 problematic exams (46.2%), which most commonly occurred during overnight shifts (48%), and often stemmed from patient positioning issues. 

  • A focus group featuring six technologist department managers led the authors to propose three additional RT resources: (1) creating ongoing training programs focused on patient positioning, (2) assigning nurses to assist technologists during exams, 3) tasking internal medicine residents with reviewing CXRs before they’re sent to radiologists.

Big Teams, Little Training – The UK’s Society of Radiographers highlighted radiography managers’ struggles with high workloads and insufficient training (n = 200), finding that many of these leaders directly manage over 20 or 30 employees (52% & 40%… yikes) and never received manager training from their hospital (45%).

  • The authors called these huge team sizes and lack of training a “gross miscalculation,” warning that it will cause managers to “undoubtedly fail in their duty of care to their staff,” especially considering that managers are often pulled into clinical duties due to understaffing.

Patient Safety’s Last Step – The WHO partnered with the ISRRT and ISR to emphasize radiographers and radtechs’ role as “the last step” in patient safety and medicine delivery, filling in gaps missed by radiation and magnetic safety experts. The collaborative webinar addressed radiographers/technologists’ responsibilities for ensuring safe contrast and radiopharmaceutical use, maintaining pediatric imaging best practices, and ensuring that pre-administration processes are complete before medication delivery. 

The Takeaway

We talk a lot about modality-based approaches to improve radtech efficiency and reduce team burnout, and those are surely needed. However, this week’s news cycle was a solid reminder of HR’s role in technologist performance and what’s at stake if techs aren’t properly supported, trained, and staffed.

Intelerad Becomes the Image Exchange Leader

Radiology took a giant step towards actually #ditchingthedisk last week with Intelerad’s acquisition of image exchange rival, Life Image. Here’s why this could be a big deal…

Exchange Leadership – Acquiring Life Image makes Intelerad the “clear medical image exchange market leader,” combining two of the top three exchange companies (the other is Nuance), and creating a far more straightforward roadmap towards building a “true nation-wide, electronic image exchange network.”

Demand & Supply – Although imaging vendors always position their acquisitions as patient or clinician-centric (even if it’s debatable), this move actually does address one of radiology’s most glaring problems — it’s far too difficult for providers to share images with each other if they don’t use the same exchange platform.

The Exchange Network Effect – Because the clinical value of image exchanges multiplies as vendor market share increases, Intelerad now has a network effect advantage that you almost never see in medical imaging. If this deal increased Intelerad’s image exchange share to 70% (hypothetically), it would make Intelerad far more valuable to its current clients and far more attractive to its remaining prospects.

Defining “Open” – The announcement alluded to the creation of an “open” image exchange, which is consistent with Ambra/Intelerad’s philosophy. However, it’s unclear how or when that will happen – or whether Nuance and other competitors will decide to join.

Intelerad = Acquirer – This deal also solidifies Intelerad’s title as imaging informatics’ most active acquirer, buying at least seven companies in the last two years that expanded it into new clinical areas (cardiac, OB/GYN), regions (UK), technologies (cloud), and functionalities (image sharing, reporting, cloud VNA). 

The Takeaway

Intelerad’s combined Ambra and Life Image acquisitions should make it the undisputed leader of the image exchange segment. That’s a big deal considering that the value of image exchange software multiplies as market share increases, and because it could actually allow Intelerad to solve (not just improve) one of radiology’s most frustrating challenges.

Echo AI Detects More Aortic Stenosis

A team of Australian researchers developed an echo AI solution that accurately assesses patients’ aortic stenosis (AS) severity levels, including many patients with severe AS who might go undetected using current methods.

The researchers trained their AI-Decision Support Algorithm (AI-DSA) using the Australian Echo Database, which features more than 1M echo exams from over 630k patients, and includes the patients’ 5-year mortality outcomes.

Using 179k echo exams from the same Australian Echo Database, the researchers found that AI-DSA detected…

  • Moderate-to-severe AS in 2,606 patients, who had a 56.2% five-year mortality rate
  • Severe AS in 4,622 patients, who had a 67.9% five-year mortality rate

Those mortality rates are far higher than the study’s remaining 171,826 patients (22.9% 5yr rate), giving the individuals that AI-DSA classified with moderate-to-severe or severe AS significantly higher odds of dying within five years (Adjusted odds ratios: 1.82 & 2.80).

AI-DSA also served as a valuable complement to current methods, as 33% of the patients that AI-DSA identified with severe AS would not have been detected using the current echo assessment guidelines. However, severe AS patients who were only flagged by the AI-DSA algorithm had similar 5-year mortality rates as patients who were flagged by both AI-DSA and the current guidelines (64.4% vs. 69.1%).

Takeaway

There’s been a lot of promising echo AI research lately, but most studies have highlighted the technology’s performance in comparison to sonographers. This new study suggests that echo AI might also help identify high-risk AS patients who wouldn’t be detected by sonographers (at least if they are using current methods), potentially steering more patients towards life-saving aortic valve replacement procedures.

Multimodal NSCLC Treatment Prediction

Memorial Sloan Kettering researchers showed that data from routine diagnostic workups (imaging, pathology, genomics) could be used to predict how patients with non-small cell lung cancer (NSCLC) will respond to immunotherapy, potentially allowing more precise and effective treatment decisions.

Immunotherapy can significantly improve outcomes for patients with advanced NSCLC, and it has already “rapidly altered” the treatment landscape. 

  • However, only ~25% of advanced NSCLC patients respond to immunotherapy, and current biomarkers used to predict response have proved to be “only modestly helpful.”  

The researchers collected baseline diagnostic data from 247 patients with advanced NSCLC, including CTs, histopathology slides, and genomic sequencing. 

  • They then had domain experts curate and annotate this data, and leveraged a computational workflow to extract patient-level features (e.g. CT radiomics), before using their DyAM model to integrate the data and predict therapy response.

Using diagnostic data from the same 247 patients, the multimodal DyAM system predicted immunotherapy response with an 0.80 AUC. 

  • That’s far higher than the current FDA-cleared predictive biomarkers – tumor mutational burden and PD-L1 immunohistochemistry score (AUCs: 0.61 & 0.73) – and all imaging approaches examined in the study (AUCs: 0.62 to 0.64).

The Takeaway

Although MSK’s multimodal immunotherapy response research is still in its very early stages and would be difficult to clinically implement, this study “represents a proof of principle” that integrating diagnostic data that is already being captured could improve treatment predictions – and treatment outcomes.

This study also adds to the recent momentum we’re seeing with multi-modal diagnostics and treatment guidance, driven by efforts from academia and highly-funded AI startups like SOPHiA GENETICS and Owkin.

CADx’s Lung Nodule Impact

A new JACR study highlighted Computer-Aided Diagnosis (CADx) AI’s ability to improve lung nodule malignancy risk classifications, while stating a solid case for the technology’s potential clinical role.

The researchers applied RevealDx’s RevealAI-Lung CADx solution to chest CTs from 963 patients with 1,331 nodules (from 2 LC screening datasets, and one incidental nodule dataset), finding that RevealAI-Lung’s malignancy risk scores (mSI) combined with Lung-RADS would significantly improve…

  • Sensitivity versus Lung-RADS-only (3 cohorts: +25%, +68%, +117%)
  • Specificity versus Lung-RADS-only (3 cohorts: +17%, +18%, +33%)

Looking specifically at the study’s NLST cohort (704 nodules), mSI+Lung-RADS would have…

  • Reclassified 94 nodules to “high risk” (formerly false-negatives)
  • Potentially diagnosed 53 patients with malignant nodules at least one year earlier
  • Reclassified 36 benign nodules to “low-risk” (formerly false-positives)

The RevealDx-based malignancy scores also achieved comparable accuracy to existing clinical risk models when used independently (AUCs: 0.89 vs. 0.86 – 0.88).

The Takeaway

These results suggest that a CADx lung nodule solution like RevealAI-Lung could significantly improve lung nodule severity assessments. Considering the clinical importance of early and accurate diagnosis of high-risk nodules and the many negatives associated with improper diagnosis of low-risk nodules (costs, efficiency, procedures, patient burden), that could be a big deal.

Viz.ai Adds PE Stratification

Viz.ai announced the FDA clearance of its new RV/LV ratio algorithm, adding an important risk stratification feature to its pulmonary embolism AI module, while representing an interesting example of how triage AI solutions might evolve.

Triage + Stratification + Coordination Viz PE becomes far more comprehensive with its new RV/LV integration, helping radiologists detect/prioritize PE cases and assess right heart strain (a major cause of PE mortality), while equipping PE response teams with more actionable information. 

  • This addition might also improve clinicians’ experience with Viz PE, noting the risk of developing AI “alert fatigue” when all severity levels are treated the same.

Viz.ai is So On-Trend – Signify Research recently forecast that AI leaders will increasingly expand into new clinical segments, enhance their current solutions, and leverage platform / marketplace strategies, as AI evolves from point solutions to comprehensive workflows. Those trends are certainly evident within Viz.ai’s recent PE strategy…

  • Viz PE’s late 2021 launch was a key step in Viz.ai’s expansion beyond neuro/stroke
  • Adding RV/LV risk stratification certainly enhances Viz PE’s detection capabilities
  • Viz PE was developed by Avicenna.AI, arguably making Viz.ai a platform vendor
  • Viz PE’s workflow now combines detection, assessment, and care coordination

The same could be said for Aidoc, which previously added Imbio’s RV/LV algorithm to its PE AI solution (and also supports incidental PE), although few other triage AI workflows are this advanced for PE or other clinical areas.

The Takeaway

Viz.ai’s PE and RV/LV integration is a great example of how detection-focused AI tools can evolve through risk/severity stratification and workflow integration — and it might prove to be a key step in Viz.ai’s expansion beyond stroke AI.

ACR Grants NPPs’ Contrast Supervision

The American College of Radiology (ACR) rolled out a significant change to its imaging contrast guidelines, allowing non-radiologists and non-physician practitioners (NPPs) to supervise intravenous CT and MRI contrast administration at accredited imaging centers.

A range of NPPs (NPs, PAs, RNs) and qualifying non-radiologist physicians will be able to directly supervise contrast administration under the “general supervision” of on-site radiologists, as long as it’s supported by state scope of practice laws. 

  • Superving radiologists must be available for “assistance or direction” and trained to handle acute contrast reactions/situations, but they won’t have to be in the same room as the patient.

These guidelines mirror the ACR’s new practice parameters for contrast supervision (adopted in May), and follow CMS’ recent efforts to expand more diagnostic tasks to non-physicians.

  • CMS granted radiology assistants the ability perform a range of imaging tasks in 2020 and permitted NPPs to directly supervise Level 2 tests in 2021 (like contrast-enhanced CT and MRI), in both cases requiring “general” radiologist supervision (on-site, but not in room… and virtual during the pandemic).

Although NPPs’ radiology expansion has historically sparked heated debates, the new ACR contrast supervision guidelines hasn’t faced many public objections so far. 

  • That’s potentially because some (busy) radiologists don’t view directly supervising contrast administration as a practical or efficient use of their time (even if they still have to drive to the imaging center), especially considering that technologists often spot adverse reactions before anyone else.
  • However, there’s surely plenty of radiologists who are concerned about whether these new guidelines might exacerbate scope creep, cut their earning potential (especially trainees), reduce radiologists’ patient-facing opportunities, and undermine patient care.

The Takeaway

The ACR’s decision to grant NPPs greater contrast supervision rights and loosen radiologists’ contrast supervision requirements might not be surprising to folks paying attention to recent ACR and CMS policies. That said, it’s still a notable step (and potential contributor) in the NPPs’ expanding role within radiology – and opinions might differ regarding whether that’s a good thing.

Prioritizing Length of Stay

A new study out of Cedars Sinai provided what might be the strongest evidence yet that imaging AI triage and prioritization tools can shorten inpatient hospitalizations, potentially bolstering AI’s economic and patient care value propositions outside of the radiology department.

The researchers analyzed patient length of stay (LOS) before and after Cedars Sinai adopted Aidoc’s triage AI solutions for intracranial hemorrhage (Nov 2017) and pulmonary embolism (Dec 2018), using 2016-2019 data from all inpatients who received noncontrast head CTs or chest CTAs.

  • ICH Results – Among Cedars Sinai’s 1,718 ICH patients (795 after ICH AI adoption), average LOS dropped by 11.9% from 10.92 to 9.62 days (vs. -5% for other head CT patients).
  • PE Results – Among Cedars Sinai’s 400 patients diagnosed with PE (170 after PE AI adoption), average LOS dropped by a massive 26.3% from 7.91 to 5.83 days (vs. +5.2% for other CCTA patients). 
  • Control Results – Control group patients with hip fractures saw smaller LOS decreases during the respective post-AI periods (-3% & -8.3%), while hospital-wide LOS seemed to trend upward (-2.5% & +10%).

The Takeaway

These results were strong enough for the authors to conclude that Cedars Sinai’s LOS improvements were likely “due to the triage software implementation.” 

Perhaps more importantly, some could also interpret these LOS reductions as evidence that Cedars Sinai’s triage AI adoption also improved its overall patient care and inpatient operating costs, given how these LOS reductions were likely achieved (faster diagnosis & treatment), the typical associations between hospital long stays and negative outcomes, and the fact that inpatient stays have a significant impact on hospital costs.

Get every issue of The Imaging Wire, delivered right to your inbox.

You might also like..

Select All

You're signed up!

It's great to have you as a reader. Check your inbox for a welcome email.

-- The Imaging Wire team

You're all set!