AI Alzheimer’s Detector | AI’s Multi-System Problem | Pipe Organ Ultrasound

“Musical instruments have a wide variety of designs but they all have one thing in common – they emit sound across a broad range of frequencies. So there is a treasure trove of design ideas for future medical imaging sensors lying waiting to be discovered amongst this vast array of designs.”

Strathclyde University professor, Tony Mulholland, on his team’s development of a pipe organ-inspired ultrasonic transducer that could help improve ultrasound image quality by broadening the range of frequencies used to emit sound waves.

 

 


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
  • OpenMarkets – A marketplace for healthcare equipment, used by hundreds of hospitals and suppliers to buy and sell imaging equipment in the most efficient way possible.
  • Pocus Systems – A new Point of Care Ultrasound startup, combining a team of POCUS veterans with next-generation technology to disrupt the industry

 

 


The Imaging Wire

 

The Early Alzheimer’s Detector
A team of researchers at UCSF, Cal, and UC Davis developed a deep learning algorithm able to diagnose Alzheimer’s disease early by identifying glucose uptake in F-FDG PET brain scans. Leveraging the Alzheimer’s Disease Neuroimaging Initiative’s (ADNI) FDG-PET dataset (2,100 images from 1,002 patients), the researchers trained the algorithm on 90% of the images and tested it on the remaining 10% as well as an independent set of 40 exams. The results were good, detecting the disease an average of over six years before final diagnosis, with 100% sensitivity and 82% specificity. The researchers suggest that when combined with other biochemical and imaging tests, this algorithm could create opportunities for physicians to slow or stop Alzheimer’s in patients.

 

AI’s Generalization Problem
Researchers from Mount Sinai Hospital and Icahn School of Medicine at Mount Sinai found that despite its promise, artificial intelligence may fall short when analyzing data across multiple health systems. The study compared a convolutional neural networks (CNN) algorithm trained on 158,000 chest X-rays across three medical institutions (Mt. Sinai, NIH, Indiana U. Hospital), finding that the algorithm was significantly less accurate diagnosing X-rays captured from an outside health system in 60% of all scans, due to inherent differences between systems (e.g. patient populations and type of CT scanner used). Given the common belief that “larger is better” when it comes to medical image training sets, this study’s finding that CNNs may have challenges generalizing data represents a new hurdle for AI to overcome. Until then, in-hospital AI ecosystems may have an accuracy advantage and multi-system CNN tools should be evaluated thoroughly before clinical use.

 

The Pipe Organ Ultrasound
Researchers at Glasgow’s University of Strathclyde developed a pipe organ-inspired ultrasonic transducer that they claim could help improve ultrasound image quality by broadening the range of frequencies used to emit sound waves. Unlike current ultrasounds that operate at a single frequency, “similar to a piano that can play just one note,” the researchers took inspiration from musical instruments (specifically pipe organs, but also piccolo) in developing an ultrasound transducer with different size/length resonating pipes, each supporting its own pitch and helping to broaden the device’s frequency range. The new transducer design is still in its very early stages, but shows promise for improving the performance of ultrasounds as well as other technologies such as hearing aids, underwater sonar, and nuclear plant testing equipment.

 

Canon’s New Celesteion PET-CT
Canon Medical Systems Japan announced the domestic launch of the Celesteion V6.5 PET-CT, touting improvements to the line’s image quality and exam time. The Celesteion V6.5’s exam time improvements are largely due to its new “Variable Bed Time” feature, which allows clinicians to apply different imaging time lengths to different parts of the body (e.g. 4 minutes for chest, 1 minute for legs), while another new feature performs chest resynchronization during the whole body scan (vs. after). The new system also improves PET image contrast with the addition of Canon’s CaLM (Clear Adaptive Low-noise Method) image reconstruction function, which reduces statistical noise while maintaining contrast. Canon certainly offers the Celesteion system in the US and Europe and it’s likely that the Celesteion V6.5 (or at least its new features) will head to western markets in 2019.

 

Q3 Financials Continue to Show Imaging Growth
The second round of medical imaging company financials from the July-September period rolled out, revealing positive financials from Fujifilm, Samsung, Shimadzu, and Siemens, but mixed results from Carestream and Hologic. Imaging related sales were strong at nearly every company.

Carestream – Onex, Carestream’s private equity parent company, increased revenue by 3% to $6.6 billion in Q2, while posting a $458 million loss (vs. $368 million in net earnings in Q3 2017). The company’s medical imaging business saw revenue decline 17% to $388 million, while net income plummeted from $558 million to a $33 million loss following the Q3 2017 sale of its dental business.

Fujifilm –Fiscal Q2 print declines drove a 1.4% drop in revenue at Fujifilm to ¥607.8 billion ($5.33b), while improvements across divisions helped operating income jump an impressive 23% to ¥47 billion ($412m, second straight quarter of big OI improvements). Fujifilm’s healthcare business scored a 11.2% revenue increase to ¥121.6 billion ($1.06b), while operating profit fell 33% to ¥1.6 billion ($14m).

Hologic – Hologic’s fiscal Q4 sales increased 1.3% to $813.5 million and net income fell 39% to $50.5 million due to a $34.8 million charge from its patent infringement lawsuit with Smith & Nephew. Hologic’s Breast Health business saw revenue increase 7.1% to $322 million during the fourth quarter. For fiscal 2018, overall revenue increased 5.2% to $3.2 billion while posting a $111.3 million net loss after achieving $755.5 million in net income last year.

Samsung – Samsung Electronics’ Q3 brought a 5.5% revenue increase to KRW 65.46 trillion ($58.3 billion) and a 20.9% increase in operating profit to a record high of KRW 17.57 trillion ($15.6 billion), helped by strong sales of its memory products and OLED panels. Samsung’s financial announcements provide little insight into the performance of its medical business, although its Consumer Electronics division (which includes healthcare) saw revenue fall 9% to KRW 10.18 trillion ($9.07 billion).

Shimadzu – Shimadzu announced a strong fiscal H1 performance, including a 6.3% increase in net sales to ¥182.85 billion ($1.6b) and a 10.7% jump in operating income to ¥17.4 billion ($152m), helped in part by its medical systems business, which posted a 6.3% increase in revenue to ¥32.6 billion ($285.9m) and a 3.1% increase in operating income to ¥685 million ($6 million). The company highlighted strong fluoroscopy sales in the US, good mobile X-ray sales in Europe, and solid angiography sales in Japan, which helped it overcome sales declines in China.

Siemens Healthineers – A “super strong” imaging performance helped Siemens Healthineers increase overall comparable revenue by 4.2% to €3.704 billion in its fiscal Q4 ($4.19b), although adjusted profit fell 70bps to €674 million (18.2% margin, $762m) and adjusted net income fell 3% to €430m ($486m). Siemens’ imaging business saw revenue grow 6% to €2.287 billion ($2.58b), while imaging adjusted profit fell 30bps to a still-solid €486m (21.2% margin, $549m), highlighting strong sales of ultrasound, CT, and X-ray products.

 

 


The Wire

  • Konica Minolta added a quintet of new and updated tools for its Exa Enterprise Imaging platform intended to improve productivity, communication, and patient care. The new tools include Exa Voice Recognition (updates Exa transcription feature with UI, customization, voice detection, and dialect and language support improvements), Exa Chat (allows users to securely communicate 1-to-1, with departments, or between hospitals and imaging centers), Exa Peer Review (built-in workflow step to share and reference patients, studies, and approved reports, eliminating need for 3rd party peer review systems), Exa Patient Kiosk (a self-service patient tool to check-in to appointments, update information, and sign electronic forms), and an improved Study Form Template (adds more flexible UI and new customization options).
  • Medical Diagnostic Web (MDW) announced the upcoming RSNA debut of its radiology blockchain platform, “designed to connect all players in the diagnostic digital imaging ecosystem and create an open, transparent and fair marketplace.” MDW claims that its platform will create business opportunities for radiologists and imaging centers while solving many current problems in imaging (service access, delivery, results communication, payment time, integration). This would allow imaging reading services to be contracted by purchasing blocks of exam interpretation credits, and then order a reading with a provider on MDW’s system, with the provider receiving payment credits via a blockchain transaction.

 

 


The Resource Wire

– This is sponsored content.

 

 

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.

Join thousands of imaging professionals.