AI is transforming radiology from the reading room to the research lab

What started as a prompt box, changed the world overnight. Instead of searching for information in a browser window, Copilot from Microsoft is empowering people to generate intelligent insights, automate tasks, and spark creativity across the applications and devices they rely on.

Generative and other types of AI are transforming radiology from the reading room to the research lab. Sometimes, generative AI looks like a more efficient meeting for the tumor board. Other times, it's a draft impression that reminds you to include details you might have missed. In all instances, it looks like empowered radiology teams, working more efficiently to improve patient outcomes.

At RSNA 2024 we showcased the expanding opportunities for radiology teams to access responsible AI, delivered at scale.¹

Giving time back to radiologists

Most radiologists still spend a significant amount of their time on tasks outside of image interpretation and report creation. From supporting a tumor board conference to drafting a grant proposal, it's important work, but it's time consuming.

It's also the kind of work that Microsoft 365 Copilot can help accelerate.

Microsoft 365 Copilot, with over 400 million users, is boosting productivity worldwide. By securely connecting and graphing the data across your applications and web activity, Copilot has the context it needs to provide more personalized and intelligent responses.

When entire radiology teams automate, search, summarize, and create with Copilot to address their everyday tasks, the minutes they save quickly add up.

In 2023, we shared how Precision Imaging Network, part of Microsoft Cloud for Healthcare, makes it easier for radiology teams to integrate third-party AI into their reporting workflows.

Precision Imaging Network offers a single point of access to a collection of third-party AI models that have gone through security, compliance, and technical reviews. There are triage AI models that work to identify imaging with critical findings and help ensure radiologists read the most urgent cases first. Detection and measurement AI models assist radiologists with image interpretation by highlighting abnormalities.

In his white paper, "A platform for healthcare enterprise AI and digital transformation," neuroradiologist and medical AI expert Dr. Reza Forghani explains how Precision Imaging Network facilitates scalable, efficient AI model deployment in weeks, not months.

"Precision Imaging Network is designed to help healthcare organizations, regardless of size, participate and lead in an AI revolution. Its framework facilitates effective and seamless integration of imaging AI models that can help them deliver better patient care."

Dr. Reza Forghani, MD, PhD, Neuroradiologist and medical AI expert

Beyond seamlessly integrating third-party AI into their workflow, radiologists can harness generative AI to help automate report creation.

PowerScribe Smart Impression, which won the AuntMinnie award for Best New Radiology Software 2024, uses generative AI to automatically create a draft impression in the radiologist's own style of dictation, using order metadata, approved AI inputs, and dictated findings. It is trained on millions of reports and delivers a range of benefits-from greater efficiency to increased job satisfaction.

Dr. Jay Janusz Kikut is Clinical Professor of Radiology and Vice Chair of Radiology Informatics at University of Vermont Medical Center. He's found that three-quarters of the radiologists in his department using PowerScribe Smart Impression believe it has made them more efficient. "It also helps them not to forget things […] and contributes to their satisfaction at work," says Dr. Kikut.

At Microsoft, we're focused on making AI available to more individuals which is why PowerScribe Smart Impression is now included with PowerScribe One.² Best of all, it's natively embedded in the workflow with no additional integrations required.

Making AI innovation safer and simpler

We're also committed to helping healthcare customers, developers, and partners build and use AI that's guided by our security, safety, and privacy principles. Our recently announced healthcare agent service in Microsoft Copilot Studio is a great example. It allows healthcare organizations to create their own generative AI-powered agents with state-of-the-art healthcare-specific chat, clinical, and compliance safeguards. This empowers them to innovate responsibly with AI. To help improve the patient experience and streamline workflows. This empowers them to innovate responsibly with AI to help improve the patient experience and streamline workflows.

Another way we're helping organizations experiment with AI is through the pretrained, multimodal healthcare AI models in Azure AI Foundry. The teams at University of Wisconsin-Madison and UW Health have been using their database of 2.5 million chest x-rays to fine-tune a model that generates draft reports with visual grounding techniques. The radiologist can click on the report content to see where the referenced areas are located directly on the chest x-ray.

"Grounded report generation from medical images is a new frontier. Collaborating with Microsoft allows us to explore the potential of these models to enhance radiology reporting and improve patient care."

Dr. Richard Bruce, Vice Chair of Informatics, Radiology at University of Wisconsin-Madison

This story is just one of many. Healthcare visionaries are applying our multimodal foundational models to address challenges from transforming radiology workflows to accelerating advancing clinical research, drug discovery, and patient care.

Supporting our partners to build a brighter future

At RSNA 2024, we shared how our collaboration with Sectra underscores our commitment to advancing healthcare through AI. By collaborating with Sectra, we are combining their leading expertise in medical imaging with our AI capabilities to develop advanced solutions that help radiologists and clinicians make faster, more informed decisions, driving improvements in healthcare outcomes.

At our booth, we showcased a stellar lineup of partners with Azure-hosted and aligned solutions. These included Flywheel's medical imaging data management platform, Volpara + Lunit's AI solution for detecting lesions suspicious for breast cancer on mammograms, and Merge by Merative's enterprise imaging workflow solution. We also featured Niramai's AI-based breast cancer screening using AI over thermal scans, NVIDIA's DGX™ Cloud serverless AI model building, and Paige's AI software for cancer diagnosis in pathology. Additionally, we shared how CitiusTech's Azure DICOM Migration Suite is using Azure Health Data Services and Microsoft Fabric to help ensure reliable data migration with high availability and disaster recovery.

A look forward to AI in radiology

AI continues to significantly transform radiology-expanding the realm of possibilities. Importantly, it can streamline workflows, enabling radiologists to focus on what matters most-improving patient care with high-quality reporting. As more radiology teams adopt and advance AI innovation, we eagerly anticipate an inspiring year ahead.

¹ Microsoft products and services (1) are not designed, intended or made available as a medical device, and (2) are not designed or intended to be a substitute for professional medical advice, diagnosis, treatment, or judgment and should not be used to replace or as a substitute for professional medical advice, diagnosis, treatment, or judgment. You are responsible for ensuring solutions comply with applicable laws and regulations.

²PowerScribe Smart Impression available for PowerScribe One versions 2019.8 and 2023.1 Service Pack 2.

Public link

Please use the above public link if you want to share this noodl on another website.