PennPET Image | Ensemble AI

“That’s just an insane amount of money, and that defeats the purpose of helping out people,”

Surgeon and imaging center owner, Dr. Gajendra Singh, on the high cost of renting an MRI, which he has to do since North Carolina won’t allow him to buy an MRI without a Certificate of Need.

Happy Thanksgiving, Imaging Wire Readers!

Heads-up that this is going to be a one-issue week due to the Thanksgiving holiday. The Imaging Wire is thankful for all of you who spend each Monday and Thursday morning catching up on the news with us and we’re super thankful for our newsletter sponsors. Please make sure to visit their sites, and if you see them at RSNA (or anywhere), thank them for sponsoring The Imaging Wire. They make this publication possible.

Keep an eye out for me at RSNA, too. I’d love to meet you.


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
  • Nuance – AI and cloud-powered technology solutions to help radiologists stay focused, move quickly, and work smarter
  • Pocus Systems – A new Point of Care Ultrasound startup, combining a team of POCUS veterans with next-generation genuine AI technology to disrupt the industry
  • Qure.ai – Making healthcare more accessible by applying deep learning to radiology imaging

The Imaging Wire



PennPET Image

A UPenn team published the first clinical images from their PennPET Explorer system, revealing that the prototype PET scanner can produce higher quality images in less time (or lower tracer doses) and with far greater versatility than current systems. Here are some details:

PennPET Explorer Tech – The PennPET Explorer uses a Philips-developed digital silicon photomultiplier, includes three rings (each with 18 detector modules), and 64cm axial FOV, while achieving 55 kcps/MBq sensitivity, 4.0 mm spatial resolution, and 250 ps time-of-flight resolution.

EXPLORER Consortium – The PennPET is the second of two large axial field-of-view (FOV) whole-body PET imagers developed by the EXPLORER Consortium, following UC Davis’ EXPLORER total-body PET/CT that debuted its first images almost exactly a year ago.

Up Next – The UPenn team is still evolving the PennPET and plans to increase its rings from three to six (expanding the axial FOV to 140 cm) and integrate a CT scanner in the future.

Ensemble AI

Brown University researchers found that using an ensemble of multiple AI algorithms targeting the same diagnostic issue is more effective than using a single algorithm, but not all ensembles work the same. Here are some details:

Ensemble Learning – The ensemble learning method combines different AI models designed to accomplish the same task, generally working best when each of the individual models performs well on their own but use different approaches and produce unique predictions.

The Study – The researchers used 48 submissions from the 2017 RSNA Pediatric Bone Age Machine Learning Challenge, creating various ensembles trained on 12,611 pediatric hand radiographs that were later tested on 200 test radiographs.

The Results – The best performing ensemble combined four models to achieve a 3.79-month estimated mean absolute deviation (MAD) bone age, beating the estimated single model MAD of 4.55 months. Interestingly, this top-performing ensemble didn’t consist of the highest-ranking individual models (that ensemble had a MAD of 3.93 months), but it did have well-performing models with the greatest diversity (0.47 pairwise correlation).


The Wire

  • Philips announced a new POCUS use case after the University of New Mexico (UNM) and Albuquerque Fire and Rescue (AFR) treated a cardiac arrest patient with the U.S.’s first out-of-hospital portable life-support system (ECMO – extracorporeal membrane oxygenation). ECMO machines have only been available within hospitals before now, but the new UNM/AFR mobile ECMO system can be used in the field, leveraging Philips’ Lumify POCUS to visually guide inserting tubes in veins and arteries.
  • Fujifilm revealed plans to preview its recently FDA-cleared FDR D-EVO III DR detector (14×17” and 17×17”) and budget-oriented FDR SE Lite retrofit DR detector at RSNA. The FDR D-EVO III was the star of this announcement, which called it the world’s first glass-free DR detector with Irradiated Side Sampling (ISS) and the world’s lightest 14×17 detector (~ 4 lbs).
  • Researchers at Austria’s St. Pölten University unveiled their SoniTalk open-source ultrasound communication protocol, which allows handheld ultrasound systems to securely transmit data to mobile phones, and reportedly offers cost and security advantages versus current methods (Bluetooth, RFID, NFC). The team plans to continue to develop SonoTalk, adding the ability to block smartphones from tracking ultrasound’s acoustic signals and add new location-tracking imaging features.
  • A recent study in the Journal of Breast Imaging found that supplemental breast cancer screening with handheld ultrasound (HHUS) or automated ultrasound (AUS) consistently improves detection of node-negative invasive cancer among women with dense breasts. The study reviewed over 400,000 screening ultrasound exams in women with dense breasts, finding that HHUS spots 2.1–2.7 cancers not seen on mammography per 1,000 women screened (90% of them invasive and node negative) and prompts an average of 7-10% of these women to undergo additional tests due to abnormalities found on screening ultrasound (~3% recommended for biopsy).
  • A North Carolina medical imaging center owner’s lawsuit to overturn the state’s MRI “certificate of need” law (CON) gained its first court victory after a Special Superior Court Judge rejected the state’s request to throw out the lawsuit. Dr. Gajendra Singh filed the lawsuit last year after the CON laws forced him to rent a mobile MRI (since he couldn’t buy his own MRI) or spend up to $500k to request a CON, calling the law unconstitutional and monopolistic.
  • Oregon-based startup Yor Labs raised $3.3 million to fund the development of its AI-enabled portable ultrasound system, after raising $1.9 million from two previous rounds. Yor Labs might not be a household name, but its leadership has a deep history in medical devices and technology (iRhythm Technologies, Cardiac Insight, Intel) and its board is well connected in the medical startup world (CareDX, DNAnexus, iRhythm, Aptus Endosystems, and others).
  • The ACR Data Science Institute outlined the four steps radiology residents and fellows should take to prepare for the growing role of AI in clinical practice. The post calls for residents and fellows to build their ability to evaluate and implement AI tools (1. Develop a Solid Foundation in Biostatistics; 2. Familiarize Yourself with ML Terminology), expand their clinical expertise as AI assumes more of their traditional responsibilities (3. Broaden Your Clinical Skillset), and consider contributing to the development of AI tools (4. Dig Deeper).
  • Siemens Healthineers announced the FDA clearance of its new SOMATOM X.cite premium single-source CT scanner (82cm bore, Vectron tube, Patient Observation Camera), highlighted by the company’s new myExam Companion intelligent UI concept that provides patient-specific prompts to help identify the optimal acquisition and reconstruction techniques for each patient.
  • An Illinois federal judge dismissed an antitrust lawsuit against the ABR that alleged it used a monopoly over the initial radiology certification process to create a monopoly over ongoing certification maintenance training. Although still not popular among many radiologists, the judge found that the ABR didn’t violate laws related to tying together multiple products because the initial certification process and certification maintenance are really two different stages of a multistage process (not different products).

The Resource Wire

  • Are you going to RSNA19? Don’t miss Dr. Irena Tocino, from Yale New Haven Health, talk about a practical guide to NLP for radiology at Nuance’s booth (#3300) at 12:30pm CT on Sunday 12/1.
  • By partnering with Medmo, imaging centers can keep their schedules full and their equipment busy. Here’s where to get started.
  • Carestream will highlight its 3D extremity imaging systems, DRX room portfolio, DRX Plus Detectors, and DR software at RSNA booth #7513.

Leave a Reply

Basic HTML is allowed. Your email address will not be published.

Subscribe to this comment feed via RSS

Get every issue of The Imaging Wire

Delivered right to your inbox. We will never sell or share your information.