Patient-Centric Portal | Open Source Imaging | Five AI Priorities

“This is the ultimate selfie.”

Applied Morphomics founder, Stewart Wang, on the level of insights available to patients through its medical imaging biomarker extraction technology.

Imaging Wire Sponsors

  • Carestream – Focused on delivering innovation that is life changing – for patients, customers, employees, communities and other stakeholders.
  • Focused Ultrasound Foundation – Accelerating the development and adoption of focused ultrasound.
  • Medmo – Helping underinsured Americans save on medical scans by connecting them to imaging providers with unfilled schedule time.
  • Pocus Systems – A new Point of Care Ultrasound startup, combining a team of POCUS veterans with next-generation technology to disrupt the industry.
  • Qure.ai – Making healthcare more affordable and accessible by applying deep learning to radiology imaging.

The Imaging Wire

Patient-Centric Portal
Research from a University of South Carolina and Cincinnati Children’s Hospital team found that EMR-based patient portal systems may be an effective way to help connect radiologists and patients. The study showed that their portal enabled patient-centric care and helped improve patients’ understanding of their medical situation, noting that many patients are interested in their radiology reports but are unable to fully understand them, even with the support of their personal physicians.

Radiologists concerned about this addition to their workflow will be relieved to know that only 0.13% of the patients in the study actually entered questions into the portal (88 out of 69k). Of those 88 patients, 47% inquired about the availability of their results, 20% sought clarifications, and 13% asked to review their images. A relatively-high proportion of radiologists (70%) in the study actively responded to questions, doing so with a 5.1-hour median response time.

Given the continued calls for greater radiologist-patient interaction and growing evidence that patients want to understand their scans and are more likely to make lifestyle changes when they do, portals like this may be a useful way to support patient-centricity without disrupting radiologist workflows.

PET May Prove CTE
New research in the New England Journal of Medicine found that Flortaucipir-PET scans combined with image analysis algorithms could be effective at detecting chronic traumatic encephalopathy (CTE) in NFL players. CTE is believed to be caused by repeated head trauma that happens while playing football, but can currently only be diagnosed post-mortem, making this potential breakthrough pretty significant.

The Flortaucipir-PET scans showed higher levels of tau protein in areas of the brain associated with CTE among former NFL players who were already displaying some CTE symptoms (n=26, SUVR 1.09-1.23) compared to a control group (n=31, SUVR 0.98-1.12). The scans also revealed higher tau levels in players who had longer football careers than players in the study who played for fewer years.

Although many are already quite certain that there’s a link between football and CTE, the ability to diagnose CTE in famous (and still-living) players would almost certainly challenge football’s role as America’s favorite sport, to play and to watch.

Five AI Priorities
An influential group of AI organizations (RSNA, ACR, NIH, and ARR) and research institutions (Stanford, Harvard Medical, Mayo Clinic, MGH, Mount Sinai, and others) just published a roadmap intended to prioritize future medical imaging AI research initiatives. Here’s the roadmap’s five steps to accelerate AI:

  1. Develop new image reconstruction methods that are suitable for human interpretation
  2. Automate image labeling and annotation methods, including info extraction, electronic phenotyping, and structured image reporting
  3. Develop new machine learning methods, including pretrained model architectures and federated machine learning methods
  4. Advance the development of explainable artificial intelligence
  5. Develop methods for image de-identification and data sharing in order to expand image dataset availability

We’ve all seen a lot of “5 steps to AI adoption”-style publications in the last few months and these ideas have been read elsewhere, but given the source of this week’s list, it’s likely that these items will indeed be prioritized in future AI research.

Open Source EIT
Imaging hardware startup, Mindseye Biomedical, just launched a crowdfunding campaign in support of its Spectra open-source medical imaging kit, targeting an initial fundraising goal of $30,000 ($14.6k raised as of yesterday) and a long-term goal of democratizing biomedical imaging. Following a successful crowdfunding campaign, Mindseye Biomedical would make the kit available to hackers and scientists for experimentation and device development, thus, “lowering the barrier of entry (into medical imaging) and opening up the biosensor playing field.”

The Spectra kit is based on electrical impedance tomography (EIT) technology, which can capture cross-section images of a range of anatomy and materials (lungs, heart, tumors, bones, prosthetics), but has not seen significant commercial or clinical adoption so far. Still, what Mindseye is doing has more to do with the open source movement than EIT technology, and although Mindseye’s path towards democratizing imaging has some obstacles (to say the least), the role of open sourced healthcare tech is almost certain to grow. This is especially true in healthcare AI, but there’s other hardware examples too.

Israel’s AI Secret
It’s not a coincidence that so many medical AI companies are Israeli, which has a smaller population than Michigan (and nine other states), but is the home of more promising healthcare AI companies than most industrialized nations. Here’s a simplified version of how Israel became an AI leader:

  • Israeli companies have a unique level of access to public health EHR data to train algorithms.
  • This access, along with Israel’s medical tech history led to a boom in Israeli digital health startups in recent years, from 327 companies in 2014 to 537 today.
  • Israeli VCs are directing more funding to healthcare startups, which increased by 32% in 2018 to $511M, 85% of which went to AI-focused firms.

Israel’s health startup trend is expected to continue, following the launch of a number of new Israeli healthcare-focused VCs over the last year and understanding that the creation of regional tech/VC clusters like this can spawn new startups and tech innovators for years (e.g. Silicon Valley, Austin, both Cambridge UK and US).

The Wire

  • New research in the American Journal of Roentgenology gives more evidence of CT colonography’s accuracy for colorectal cancer screening, while revealing additional risk stratification benefits compared to stool-based tests. The study looked at 1,650 patients who underwent CTC exams followed by colonoscopy, finding that CTC had a positive predictive value of 90.8% by-patient and 88.8% by-lesion for lesions 6mm and larger (vs. colonoscopy’s 100%). CTC was also effective at specifying the nature of positive findings, achieving a 72.3% patient PPV for neoplasia and 38.8% for advanced neoplasia lesions larger than 6mm.

  • EOS imaging unveiled the company’s new EOSlink solution, which integrates its EOSapps preoperative surgical planning software (spine, hip, and knee surgeries) with intraoperative surgical solutions (e.g. navigation devices, robotics systems, and custom spinal rod solutions). EOS Imaging’s launch of EOSlink and the still-new EOSapps are part of its effort to provide a full suite of orthopedic hardware and software solutions.

  • Relatively large US Imaging center company, Akumin (now an estimated 117 centers), signed a deal to acquire 27 imaging centers in Florida and Georgia from Advanced Diagnostic Group for roughly $214 million. The deal significantly expands Akumin’s personal injury imaging business, adds 21 centers to its already large presence in Florida (now ~66), and expands the company to Georgia for the first time with four centers (soon to be 6).

  • Israeli imaging AI company, Aidoc, completed a pretty significant $27 million Series B round that it will use to fund its ongoing R&D and commercialization efforts, revealing a goal to grow its presence from 100 hospitals today to 500 sites in two years. With $40 million raised to-date, Aidoc is now among the better-funded medical imaging AI companies, behind super-funded AI companies like HeartFlow ($476.6m) and VoxelCloud (~$80m) and in the same ballpark as Zebra Medical (~$50m) and Arterys ($43.7m).

  • UCLA researchers developed an AI system that can use MRI scans to identify and predict the aggressiveness of prostate cancer with nearly the same accuracy as experienced radiologists. UCLA’s FocalNet artificial neural network system was trained with MRI scans from 417 men with prostate cancer, later detecting prostate cancer with 89.7% and 87.9% sensitivity (index lesions and clinically significant lesions), which was statistically comparable to radiologists with at least 10 years of experience (93.1%, 89.4%). FocalNet also assessed cancer aggressiveness with AUCs of 0.81 and 0.79. The researchers noted the years of training required to gain expertise in prostate cancer MRI diagnosis, suggesting that algorithms like this may reduce training requirements or allow less-experienced radiologists to diagnose prostate cancer.

  • NPR continued its almost weekly coverage of healthcare AI, this time with a Weekend Edition piece questioning its safety. Weekend Edition detailed the challenges that AI creates for the FDA (AI is generally tested with specific patient populations, can be self-learning, etc.), many of which haven’t been solved, and suggested that there may be situations where clinicians may not know what AI-driven diagnoses are safe to believe. Once again, this isn’t really new news for folks in radiology or health tech, but it’s definitely shaping the opinions of Weekend Edition’s 4 million listeners.

The Resource Wire

This is sponsored content.

  • Yale University research reveals that the average patient drives past SIX lower-priced providers on the way to an imaging procedure, due in large part to patients’ and physicians’ limited cost consciousness. Medmo helps address this issue by letting patients enter the price they can afford for their scan, then books them at a nearby imaging center willing to accept that price.

  • POCUS Systems has developed a low-cost, hi-resolution, and AI-enabled POCUS unit, designed to diagnose specific MSK abnormalities and eliminate cost, size, and training barriers.

You might also like

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!