Google’s Imaging AI | Breast Cancer Predictor | Watson Imaging Goes Live

“The blood vessel segments are the streets and the blood flow in each segment is analogous to the traffic along each street.”


Johns Hopkins associate professor of radiology and biomedical engineering, Arvind Pathak, PhD, on a “Google Maps” approach his team developed to help visualize the blood vessel changes associated with tumor growth.


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 technology to disrupt the industry.
  • Qure.ai – Making healthcare more accessible by applying deep learning to radiology imaging.

Imaging Wire Sponsors


Google’s Imaging AI
Medical imaging AI played a prominent role in this week’s Google I/O developer conference, where the search giant promoted a Google Brain algorithm trained to find early signs of cancer in CT scans that could improve survival rates by 40%. The algorithm, built through a partnership with the National Cancer Institute and Northwestern University, was specifically highlighted for spotting a patient’s early-stage cancer that was missed by 5 out of 6 radiologists.

Calling this a “a promising, but early result,” Google revealed plans for more partnerships with the medical community as it tries to advance this and other algorithms. Plans like that can certainly get the attention of the healthcare industry, especially given the growing belief that Silicon Valley is about to invade medicine, and considering that Google’s AI and Cloud businesses have consistently targeted radiology in recent years.

However, it’s worth noting that the rest of Google I/O 2019 was devoted to the company’s hardware gadgets and online widgets, and even if it pushed further into healthcare, Google is much more likely to serve as an AI technology provider than an end-to-end medical AI company.


MIT’s Breast Cancer Predictor
Researchers from MIT and MGH developed a mammography-based deep learning breast cancer risk model that’s more accurate than established clinical breast cancer models based on density and family history and performs similarly across different ethnic groups.

The team used 71,689 images to develop a trio of DL risk models (a traditional risk factor model, an image-only model, a hybrid model combining DM images and risk factors) and then compared them to the Tyrer-Cuzick breast density risk model. The hybrid DL model proved to be the most accurate (0.70 AUC, 31% of patients who later developed breast cancer ID’d as high-risk), with a statistically significant advantage over the Tyrer-Cuzick model (0.62 AUC, 18% ID’d as high-risk). The hybrid DL model was particularly accurate at predicting breast cancer among African American women, achieving the same 0.71 AUC as white women, which is notable given the Tyrer-Cuzick model’s significant racial variability (0.62 AUC white, 0.45 AUC African American).

Although more research is needed, this study provides strong evidence that mammography may be able to support risk prediction, potentially replacing conventional density-based risk models and allowing earlier breast cancer detection. Given the role of density in breast health today, that would be a really big change.


Watson Imaging Goes Live
Kentucky’s Hardin Memorial Health (HMH) became the first hospital system to go live with IBM Watson Imaging Patient Synopsis, effectively making HMH the first provider “to put Watson to use in medical imaging.” This is a pretty major milestone given the prominent roles that IBM Watson and medical imaging have respectively played in healthcare AI’s short history.

Patient Synopsis supports radiologists while they’re reading imaging studies, using radiologist-trained AI to extract patient information from EHRs and provide key insights to radiologists within a “single-view summary.” HMH, a 15-year IBM Merge Unity PACS client, will serve as a test location for Patient Synopsis, presumably before a more widespread launch.

In addition to representing a big and long-awaited achievement for IBM, this could be the starting point of a larger push into radiology by Watson Imaging (beyond IBM’s PACS/EI business). The next step may come from Watson Imaging Clinical Review (spots discrepancies between diagnostic reports and EHR problem lists), which has been scheduled to go live at Ohio’s TriHealth in 2019, while the future could bring an expansion beyond the EHR with new solutions that “leverage Watson image-analysis capabilities.” Still it’s probably wise not to look too far into the future and appreciate this announcement for what it is – the first time Watson has been used in medical imaging.


The Wire

  • Doctors in Guangzhou, China conducted an ultrasound scan on a patient from 37 miles away by using 5G wireless technology to remotely control a robotic arm holding the ultrasound scanner. The physicians highlighted 5G technology’s ability to solve previous issues regarding audio and video time lag, revealing plans to expand 5G to more medical applications in the future (e.g. consultations, surgery instructions, emergency response).

  • GE Healthcare and Indian tech innovation hub, NASSCOM, launched a partnership with the goal of co-developing digital health solutions for the Indian market. GE will work with Indian startups through NASSCOM’s Center of Excellence-Internet of Things (CoE-IoT) to co-creating a range of digital applications, while also working with the Indian government to help shape the country’s digital health policies. GE Healthcare is the latest of a number of major companies to partner with NASSCOM, which apparently has similar partnerships with Intel, IBM, Qualcomm, and Microsoft, among others.

  • The Indian city of Chennai has reportedly experienced significant drops in imaging scan costs (CT and MRI scans down 30% since 2016), due to rising competition and efforts to capture unrealized demand from the city’s lower-income patients. Chennai’s radiology cost declines began after public hospitals expanded their imaging fleets/capacity and reduced their fees, prompting private centers to react with their own rate reductions and adopt some unique promotional strategies including reducing rates during evening hours or for appointments that are made using specific radiology booking apps.

  • Carestream expanded its ImageView Software to its DRX-Revolution Mobile X-ray System, after initially supporting the company’s OnSight 3D Extremity Imaging System, revealing plans to make the software available across its entire portfolio. ImageView supports image processing and radiographer workflow, combining visualization processing, tube and line visualization, pneumothorax visualization, bone suppression software, pediatric image optimization / enhancement software, SmartGrid software, and access to RIS and PACS platforms.


The Resource Wire

This is sponsored content.

  • This Qure.ai blog details the challenges it overcame with its qER Head CT algorithm, including developing the algorithm to support CT’s 3D images and high resolution as well as the validation hurdles it faced due to the low prevalence of abnormalities, the need to create a dataset enriched with positives, and steps required to support radiologist reading.

  • How much does a CT scan cost? According to Medmo, that depends. Scans made with the exact same device on the exact same body part could cost $225 at one facility and $2,500 at another. Medmo also provides some advice to make sure patients don’t pay too much for their scans, including using the Medmo Marketplace, where the average CT costs between $225 and $700.

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.