Stay organized with collections Save and categorize content based on your preferences.

Johns Hopkins University BIOS Division: Advancing intracerebral hemorrhage treatments through AI

About Johns Hopkins University BIOS Division

The Johns Hopkins University (JHU) Brain Injury Outcomes (BIOS) Division is dedicated to improving interventions through data-science-driven clinical trials and medical image reviews.

Industries: Healthcare, Life Sciences
Location: United States

Tell us your challenge. We're here to help.

Contact us

About Quantiphi

Quantiphi, a premier Google Cloud Partner, is an award-winning applied AI and data science software and services company driven by the desire to solve transformational problems at the heart of business.

The JHU BIOS Division has been working on medical imaging to enable faster and more accurate decision-making for brain injury patients. Google Cloud and Quantiphi supported breakthroughs in their efforts through cloud-based machine learning services.

Google Cloud results

  • Accelerates generating insights from scans from approximately 500 patients from 2,500 hours to 90 minutes
  • Enhances accuracy of brain scan-related data points by dice coefficient of 0.93
  • Supports faster experimentation, thanks to higher processing speeds in the cloud
  • Makes trial core laboratory functions a real-time capability
  • Reduces research and infrastructure costs
  • Enables integration of multiple small fragments using Healthcare API

Reduces clinical trial-related brain scan review times from five hours to 30 seconds

Strokes are the fifth leading cause of deaths in the U.S. The U.S. Centers for Disease Control and Prevention reports that roughly 800,000 Americans suffer a stroke each year; 140,000 of them die from it. To combat this alarming trend, researchers have been striving to develop new treatments, accelerate delivery of care to patients, and improve the quality of life for stroke survivors.

Johns Hopkins University’s Brain Injury Outcomes (BIOS) Division has focused on studying brain hemorrhage in the hopes of unlocking new paths of treatment that ultimately improve medical outcomes following a stroke. The team has several decades of experience in developing and improving models for analyzing brain images of hemorrhage patients, including quantifying volumes and other risk factors of brain hemorrhage through patient CT scans.

Two years ago, the BIOS Division gained the support of the National Institutes of Health (NIH) Science and Technology Research Infrastructure for Discovery, Experimentation, and Sustainability (STRIDES) initiative. The initiative uses Google Cloud as its technology foundation. As a result, the BIOS team began to use Google Cloud solutions, including Compute Engine and Cloud Dataflow, as well as analytics support from partner Quantiphi, to accelerate its research.

“The AI-driven approach is a much more robust method than our previous more costly and time-consuming manual approach to analyzing images. It could help us apply insights to a patient in as little as 30 seconds. This could result in bringing an NIH study-quality core lab, which has taken 30 years to develop, to every American who has an accessible digital imaging file.”

Daniel F. Hanley, Jr., M.D., Director, Johns Hopkins University BIOS Division

Rapid insights through big data, machine learning

Daniel F. Hanley, Jr., M.D., a professor of neurology at Johns Hopkins University and Director of the BIOS Division, recognized the value of advanced analytics and machine learning solutions years ago. Using the massive database provided by STRIDES, Dr. Hanley and his team began to convert CT scans into 3D images—priming this data for powerful analysis that would ultimately provide a more accurate perspective on blood clots in the brain.

“We used the world’s best-curated library of brain images and brain hemorrhage patients to develop algorithms and train machine learning systems to dramatically reduce the amount of time it takes to evaluate scanned imagery,” says Dr. Hanley.

Traditionally, the time commitment to review a sufficient number of brain scans—from roughly 500 patients—would be about 2,500 hours spread across 7–10 researchers. With the new AI algorithm powered by Quantiphi, Google Cloud Machine Learning Partner of the Year, and Google Cloud, Dr. Hanley and his team reduced the time per scan from five hours to 30 seconds. This translated to a 4.5 hour evaluation time versus 2,500 hours.

On top of the notable time savings, the BIOS Division found that using machine learning to support the analysis was more accurate than having people review and analyze the same images.

“We use 3D Unet architectures across a variety of images that we trained the model to recognize and have achieved a dice coefficient of .93 that is proportionate to the overlap of an expert compared to the machine,” says Asif Hasan, co-founder of Quantiphi. “This is state of the art.”

While the research is helping to illuminate paths forward by way of treatments, the biggest impact comes at the point of care. “The AI-driven approach is a much more robust method than our previous more costly and time-consuming manual approach to analyzing images. It could help us apply insights to a patient in as little as 30 seconds,” says Dr. Hanley. “This could democratize the NIH study-quality core lab, which has taken 30 years to develop, by bringing it to every American who has an accessible digital imaging file.”

“We’ve aligned closely with the goal of showing that a cloud-based, AI-driven approach is robust and that it can be performed while protecting personal PHI. In terms of costs, the cloud has definitely reduced some of the traditional financial demands of our research.”

Daniel F. Hanley, Jr., M.D., Director, Johns Hopkins University BIOS Division

Bringing clinical research safely to the cloud

Dr. Hanley and his team have found that, while cloud technology’s impact on the research itself is valuable beyond many metrics, it may also help drive costs down by 50% or alternatively enable more robust screening and core laboratory oversight by uniformly utilizing machine learning.

“We’ve aligned closely with the goal of showing that a cloud-based, AI-driven approach is robust and that it can be performed while protecting personal PHI,” says Dr. Hanley. “In terms of costs, the cloud has definitely reduced some of the traditional financial demands of our research.”

Using Google Cloud has enabled both Quantiphi and the BIOS Division to experiment with a range of cloud computing solutions. Dataflow and Cloud Healthcare API have assisted with some of the integration of digital imaging and communications in medicine (DICOM) files, while the team has used Compute Engine for experimentation, AI Platform for distributed training to accelerate iterations, and Google Kubernetes Engine to automate the inference pipeline. The integrated approach works to feed data into state-of-the-art machine learning models that generate the valuable insights that have the potential to improve health outcomes.

As more hospitals and laboratories begin to migrate to the cloud, the opportunities to integrate the vast treasure troves of data for research purposes will expand.

“Core laboratories in ischemic stroke are now migrating to both hospital-based servers and, in the last year, some of the information management and knowledge transfer about the core lab findings are migrating to cloud-based platforms,” says Dr. Hanley. “We’re going to see that happen in brain hemorrhage with the excellent work that Quantiphi has been doing.”

“There are thousands of quantitative measures of brain images that can be used to improve diagnosis, treatment, and outcomes in any brain injury—and many of these measures likely apply to every other organ. It used to be burdensome and time intensive to gain this level of insight, but now using more advanced technology to facilitate our work is truly transformative.”

Daniel F. Hanley, Jr., M.D., Director, Johns Hopkins University BIOS Division

Partnering for better patient outcomes

The combination of the STRIDES program’s robust database of medical images, Quantiphi’s mastery of machine learning algorithms, and the power of Google Cloud have helped Dr. Hanley and his team accelerate health outcomes, all while keeping costs and workloads in check.

Using Google Cloud technology, Quantiphi’s expertise, and the growing volume of medical data, Dr. Hanley and his team intend to prove the viability, potential, and promise of this approach to accelerate research to improve healthcare decision-making within the field of brain injuries. But the benefits of their work won’t end there.

“There are thousands of quantitative measures of brain images that can be used to improve diagnosis, treatment, and outcomes in any brain injury—and many of these measures likely apply to every other organ,” says Dr. Hanley. “It used to be burdensome and time intensive to gain this level of insight, but now using more advanced technology to facilitate our work is truly transformative.”

Tell us your challenge. We're here to help.

Contact us

About Johns Hopkins University BIOS Division

The Johns Hopkins University (JHU) Brain Injury Outcomes (BIOS) Division is dedicated to improving interventions through data-science-driven clinical trials and medical image reviews.

Industries: Healthcare, Life Sciences
Location: United States

About Quantiphi

Quantiphi, a premier Google Cloud Partner, is an award-winning applied AI and data science software and services company driven by the desire to solve transformational problems at the heart of business.