Technology

Neurology CT

Cardiac CT

Women's Health CT

Oncology CT

Facility Management

Neurology MRI

Oncology MRI

Women's Health MRI

Cardiac MRI

Mammography

Women's Health Ultrasound

Oncology Ultrasound

Cardiac Ultrasound

Ultrasound

Orthopedic X-Ray

X-Ray

News

All News

Blogs

Media

Around the Practice 

Between the Lines

Image IQ

Medical World News

Podcasts

Videos

Conferences

Conference Coverage

Conference Listing

DI Executive

Resources

CME/CE

E-books

Education

Partners

Subscribe

Advertise

Contact Us

Editorial

Editorial Board

Do Not Sell My Personal Information 

Privacy

Terms and Conditions

Diagnostic Imaging

Facebook

Twitter

LinkedIn

YouTube

Spotlight

Disease Spotlight: Prostate Cancer

Image IQ Quiz

RSNA Coverage

Topics

View More

1rem

Topic

 AI (artificial intelligence) focus page provides information, videos, podcasts, and the latest news about product developments, trial results, screening guidelines, and protocol guidance that touch on the development and use of AI across the healthcare continuum.

Artificial intelligence has been a mainstay of the Radiological Society of North America (RSNA) annual meeting for nearly a decade. Each year brings new developments, reveals new capabilities and algorithms, and furthers the conversation about how these tools can help radiologists provide care.

This year, the conversation began to turn to what steps could come next for AI. Conversations about deployment, consolidation, and U.S. Food & Drug Administration clearance are now the norm as the technology becomes less of a novelty and more of a mainstay.

 speaks with Nina Kottler, M.D., chief medical officer of AI at Radiology Partners, about her reaction to what this year’s conference offered about these tools and what she sees on the horizon in 2021.

RSNA

Nina Kottler, M.D., chief medical officer of AI at Radiology Partners, discusses, during RSNA 2020, what new developments the annual meeting provided about these technologies, sessions to access, and what to expect in the coming year.

The Reading Room: Artificial Intelligence: What RSNA 2020 Offered, and What 2021 Could Bring

Five artificial intelligence (AI) algorithms for mammography assessment were better at predicting breast cancer risk over five years than the Breast Cancer Surveillance Consortium (BCSC) risk model, according to new retrospective research involving over 13,000 women.

Study Assesses Ability of Mammography AI Algorithms to Predict Breast Cancer Risk

Powered by deep learning technology, Sonic DL reportedly facilitates the acquisition of cardiac magnetic resonance imaging (MRI) scans at 12 times the speed of conventional MRI systems.

FDA Clears Cardiac AI Software That May Reduce MRI Scan Times by 83 Percent

The artificial intelligence (AI)-powered Ezra Flash reportedly enhances magnetic resonance imaging (MRI) and enables significant reductions in scan times and costs for full-body MRI.

FDA Clears AI Software That May Lead to 30-Minute Full-Body MRI Exams

Precision DL, a deep learning-based software which will be available on GE HealthCare’s Omni Legend PET/CT device, reportedly increases the detectability of small, low-contrast lesions by 42 percent.

FDA Clears New AI-Powered Imaging Processing Software for PET/CT

Emerging breast ultrasound research showed the use of computer-aided diagnosis (CAD), powered by deep learning, led to 24 percent and 36.9 percent improvements in accuracy and specificity, respectively, in the use of BI-RADS classifications by radiologists without breast ultrasound expertise.

Multicenter Breast Ultrasound Study: AI Bolsters Accuracy and Specificity of BI-RADS Classifications

The recently released ChatGPT-4 (Open AI) may offer more advanced reasoning, be less prone to hallucinations, and be more capable of passing a radiology board exam than ChatGPT-3.5 (Open AI), according to newly published research.

In prospective studies published recently in 

, researchers assessed the performance of 

 in answering 150 text-based multiple-choice questions akin to those found on a radiology board examination.

The researchers found that the ChatGPT-4 model correctly answered more than 80 percent of the questions in comparison to 69 percent for ChatGPT-3.5. ChatGPT-4 also demonstrated a greater than 20 percent improvement over ChatGPT-3.5 on questions that required higher-order thinking, including description of imaging findings, classifications, and application of concepts, according to the study authors.

In a recent interview, Rajesh Bhayana, MD, FRCPC, the lead author of the studies, said it is apparent that the technology with ChatGPT is showing significant improvement.

“The fact that ChatGPT-4 performed better than ChatGPT-3.5 and had less frequent incorrect answers and also performed better with higher-order reasoning suggests the frequency of hallucinations is in fact decreasing,” noted Dr. Bhayana, an abdominal radiologist, and technology lead in the Department of Medical Imaging at the University of Toronto in Canada.

(Editor’s note: For related content, see “

Can ChatGPT Provide Appropriate Information on Mammography and Other Breast Cancer Screening Topics?

While Dr. Bhayana said there is significant potential with the use of ChatGPT in radiology, he cautioned that accuracy remains an issue and use of the technology still requires rigorous fact checking.

“It was very impressive that these models, based on the way they work and based on the fact that they are general models, performed so well in a specialty like radiology where language is so critical,” maintained Dr. Bhayana. “(But) it still does get things wrong. When it does get those things wrong, it uses very confident language. If you’re a novice and you can’t separate fact from fiction, it can be tough to know what’s right and what’s wrong. Especially for education, especially for novices when you’re looking up that information and learning something for the first time, you can’t rely on it. If you do use it, you have to always fact check it.”

For more insights from Dr. Bhayana, watch the video below.

In a recent interview, Rajesh Bhayana, M.D., shared insights from new research that compared the abilities of ChatGPT-3.5 and ChatGPT-4 to answer text-based questions akin to those found on a radiology board examination.

Can ChatGPT Pass a Radiology Board Exam?

When evaluating an artificial intelligence (AI) model for initial assessment of left ventricular ejection fraction (LVEF) on echocardiograms, David Ouyang, M.D., said he was genuinely surprised that cardiologists had greater confidence in initial AI assessments than those provided by cardiac sonographers with an average of 14.1 years of experience.

In the randomized trial, recently published in 

, cardiologists revised 27.2 percent of initial LVEF assessments by sonographers in comparison to 16.8 percent of AI assessments.

“We initially thought of the tool as kind of a streamlining tool where it speeds things up. We actually designed the trial as a non-inferior trial. We weren’t expecting AI to be better than sonographers. Hopefully, it would be equivalent,” noted Dr. Ouyang, who is affiliated with the Department of Cardiology at the Smidt Heart Institute and the Division of Artificial Intelligence in Medicine at the Cedars-Sinai Medical Center in Los Angeles. “We were pleasantly surprised that AI actually was superior. This really speaks to the high level of precision that automation delivers.”

Study Finds AI More Effective Than Sonographer Interpretation of Cardiac Function on Echocardiograms

Can AI Improve Triage of CT Pulmonary Angiography Exams for Acute PE?

AI Platform for Ultrasound Detection of Cardiac Amyloidosis Gets FDA Breakthrough Device Designation

In a recent interview, Dr. Ouyang also discussed another 

 that employed deep learning segmentation of cardiac magnetic resonance imaging (MRI) to examine how variations in left ventricle sphericity can impact cardiovascular risks. Based on their findings from a large cohort of nearly 39,000 participants from the UK Biobank, Dr. Ouyang and colleagues found that one standard deviation increase in the left ventricle sphericity index led to a 47 percent increased risk for cardiomyopathy and a 20 percent higher risk of atrial fibrillation.

For more insights from Dr, Ouyang, watch the video below.

In a recent video interview, David Ouyang, M.D., shared insights from two recent studies he co-authored on the use of artificial intelligence (AI) to improve initial assessment of left ventricular ejection fraction (LVEF) on echocardiography and ascertain cardiac risks associated with changes in the left ventricle sphericity index seen on magnetic resonance imaging (MRI).

Emerging AI Advances in Cardiac Imaging

ChatGPT may have considerable potential for streamlining radiology workflows and reporting but concerns over optimal prompt engineering and hallucination effects may curtail the use of the artificial intelligence (AI) chatbot in radiology for the time being.

In recent video interviews. Nina Kottler, M.D., MS, Tessa Cook, M.D., Ph.D, and Sonia Gupta, M.D., discussed the potential capabilities of ChatGPT for radiologists and cautioned about current drawbacks of the technology.

Dr. Kottler suggested that ChatGPT could help remedy the frustration of receiving incomplete patient histories.

“We’ve never (received) great (clinical) histories. That decreases the quality of our output and causes us to hedge a lot,” notes Dr. Kottler, associate chief medical officer of Clinical AI and vice-president of Clinical Operations at Radiology Partners. “ChatGPT could either summarize information from the EMR (electronic medical record) and provide us with that robust patient history or summarize prior reports for us so we get that history and the pathology that we’re finding.”

Another potential benefit of ChatGPT could be converting patient summaries of radiology reports into different languages and possibly expanding access to care, suggested Dr. Gupta, the chief medical officer of Change Healthcare and a board member of the American Board of Artificial Intelligence.

Emerging Insights on Improving Radiology Workflows

Assessing the Value Proposition of AI in Radiology

Emerging Trends with AI in Radiology in 2023

However, Dr. Kottler noted that the quality of the prompt input has a strong effect on the results one gets with ChatGPT. Drs. Cook, Gupta and Kottler all emphasize having appropriate safeguards in place to ensure the accuracy and verifiable referencing of information provided by ChatGPT. While Dr. Cook noted use cases with ChatGPT that have provided a synthesis of literature on a given topic, she said other cases have demonstrated “hallucination effects” in which there has been rewritten history or the “invention of citations that don’t exist.”

“Just because ChatGPT is far more fluent than any other chatbot we might have encountered in the past doesn’t necessarily mean it is not fallible in some way as well,” cautioned Dr. Cook, an associate professor of radiology, and director of the Center for Practice Transformation in Radiology at the Hospital of the University of Pennsylvania.

For more insights from Drs. Cook, Gupta and Kottler, watch the video below.

In recent video interviews, Tessa Cook, MD, PhD, Nina Kottler, MD, MS, and Sonia Gupta, MD shared their insights and perspective on potential benefits and drawbacks of ChatGPT in radiology.

Can ChatGPT Have an Impact in Radiology?

One of the ongoing trends with the development of artificial intelligence (AI) modalities in radiology is an increasing emphasis on improving efficiencies with non-interpretative tasks.

In a recent interview, Sonia Gupta, MD said the combination of AI and natural language processing can help retrieve and analyze key historical points from a patient’s electronic medical record (EMR). Dr. Gupta emphasized that this technology could help streamline patient triage, radiology reporting and the management of incidental finding follow-up.

“(With) incidental finding follow-up, AI reads the report and detects the findings that need follow-up … and can also insert the follow-up recommendations for you into the report. Again, this is all non-interpretative and natural language processing-based type of AI. I think that is a really great opportunity we can all utilize,” maintained Dr. Gupta, an abdominal radiologist, a board member of the American Board of Artificial Intelligence in Medicine, and the chief medical officer of Change Healthcare.

AI in Radiology: Top Five Articles of 2022

Dr. Gupta said the hope is that this emphasis on simplifying non-interpretative tasks could reduce time-consuming elements of radiology workflows and help curtail burnout.

Noting that approximately one-third of radiologists are currently utilizing some form of AI in practice, Dr. Gupta said the combination of cancer screening with AI and appropriate reimbursement could accelerate AI adoption in radiology.

Pointing out there are currently upward of 200 Food and Drug Administration (FDA) 510(k) clearances in radiology, Dr. Gupta said the ongoing consolidation of AI vendors may make it easier for radiologists to choose among relevant systems for their practice. She added that consolidation could be a boon for early adopters, who may gain access to more AI offerings than they originally signed up for.

For more insights from Dr. Gupta, watch the video below.

In a recent video interview, Sonia Gupta, MD discussed a number of ongoing developments with artificial intelligence (AI) in radiology, ranging from market consolidation of AI vendors to maximizing automation and efficiency with patient triage, reporting and follow-up of incidental findings.

Combining artificial intelligence detection (AI) algorithms with radiologist assessment may improve detection rates anywhere between two to 45 percent, according to Nina Kottler, M.D., M.S., the Associate Chief Medical Officer of Clinical AI and Vice-President of Clinical Operations at Radiology Partners.

However, in a recent video interview, Dr. Kottler emphasized that recognition and appropriate education on the potential biases of AI algorithms and inherent radiologist biases are key to maximizing the impact of AI on patient outcomes.

Dr. Kottler noted that factors such as the patient history, the exam type ordered, an unusual location for a lesion and distracting pathology can contribute to variability with radiologist assessment. She also pointed out that education biases stemming from one’s experience, recently missed diagnoses and satisfaction of search can also factor into image interpretation.

However, Dr. Kottler added there are also biases or limitations with the development of AI algorithms as well.

“Most AI systems are trained on one series or a couple of images within a total image set,” noted Dr. Kottler, who lectured about AI at the recent Radiological Society of North America (RSNA) conference. “If you’re looking at a chest X-ray, a lot of the chest algorithms are trained on just the frontal image. If you’re looking at a head CT for intracranial hemorrhage, most of those (AI) algorithms are trained just on the axial, soft tissue, thinnest slice series with no contrast. It's not looking at all of the other components.”

(Editor's note: For other video interviews with RSNA lecturers, click 

Dr. Kottler emphasized appropriate education that goes beyond the sensitivity rates and area under the curve (AUC) performance of AI systems.

“We spend a lot of time in our practice trying to understand the bias of the AI (model)," maintained Dr. Kottler. "Where is it overcalling things? Where is it under-calling things? We take that information to our radiologists and let them know. We just don’t give them information like (this AI system) has a 95 percent AUC, it is going to be great. No. We go to them and say (this system) may overcall intracranial hemorrhage when the patient is moving and there is a streak artifact in the exam. It’s going to miss some of the findings in X, Y and Z cases. You bring those things to the radiologists and you’re educating them on the biases of the AI. You’re making it so (radiology) and AI are stronger together.”

For more insights from Dr. Kottler, watch the video below.

In a video interview discussing one of her recent lectures at the Radiological Society for North America (RSNA) conference, Nina Kottler, M.D., M.S., noted how the combination of artificial intelligence (AI) and radiologist experience can help mitigate bias limitations with the development of AI algorithms as well as educational biases inherent to a radiologist’s training and experience.

Recognizing and Addressing Biases with AI and Radiologists

While there are varied perspectives on what “the cloud” represents for radiologists, Eliot Siegel, M.D., sees the promise of increased efficiency and improved patient care as the field continues to grapple with increasing imaging volume and radiologist shortages.

Beyond off-site storage of medical imaging data, cloud-based resources can provide immediate access to timely PACS updates and alternative backup options, and potentially serve as a centralized hub that may help streamline access to a patient’s clinical history and past imaging from other institutions, noted Dr. Siegel during an interview at the recent 

Radiological Society of North America (RSNA) conference

“The cloud represents a central repository that could be more patient-centered rather than institution-centered. That is a major advance,” said Dr. Siegel, a professor of Diagnostic Radiology and Nuclear Medicine and vice-chair of Information Systems at the University of Maryland School of Medicine.

Dr. Siegel also noted that cloud-based image management could enhance access to a variety of artificial intelligence (AI) platforms.

“Many of the AI vendors are offering software that is really interesting and can help me become safer, more efficient, more effective, and more productive. Many of those are things I need to consume in the cloud,” noted Dr. Siegel. “The potential of being able to send my images up to a site in the cloud and then be able to have all those different AI services and other services that I want to be able to directly interface with in the cloud is extraordinarily compelling.”

In conversations Dr. Siegel has had with information technology (IT) specialists and chief information officers (CIOs), he said he doesn’t hear concerns about data security that he has heard in years past about the cloud. While cost may be the biggest concern expressed about implementation of cloud-based image management, Dr. Siegel said there are considerable costs with local image storage, and it may get to an inflection point where economies of scale become advantageous with companies that provide cloud-based services.

For more insights from Dr. Siegel, watch the video below.

In a recent interview at the Radiological Society of North America (RSNA) conference, Eliot Siegel, M.D., discussed a variety of potential benefits with cloud-based image management in radiology, ranging from enhanced data security and economies of scale to improved access to a variety of artificial intelligence (AI) solutions to increase efficiency.

Maximizing Cloud-Based Capabilities in Radiology

While increased imaging volume remains a daunting challenge in radiology, emerging developments in artificial intelligence (AI) and natural language processing (NLP) may help alleviate worklist fatigue.

In an interview from the recent RSNA conference, Tessa Cook, M.D.,Ph.D. spoke about the potential of NLP to assist with imaging protocols in radiology, the capability of AI to streamline radiology workflows and recently published research on an emerging model for automating de-identification in radiology reports.

Along with colleagues from the University of Pennsylvania and Stanford University, Dr. Cook co-authored a recently published study assessing a new model that provides automated de-identification for radiology reports.

While she noted an abundance of de-identification tools for DICOM data and the likelihood of that capability being built into PACS for many radiologists, Dr. Cook said this is more challenging for protected health information (PHI) in radiology reports.

“The distribution of (PHI) data is very different for radiology reports than it is for medical text,” said Dr. Cook, an Associate Professor of Radiology, Chief of 3-D and Advanced Imaging and Director of the Center for Practice Transformation in Radiology at the Hospital of the University of Pennsylvania.

Dr. Cook noted that NLP could help ease the time-consuming nature of assigning imaging protocols given the existence of large quantities of training data that can capture the homogeneity of an imaging practice. She added that AI could be beneficial for more quantitative tasks such as comparisons and the tracking of lesion measurements over time.

“There are so many use cases (for AI). I think the ones that are going to succeed are the ones that are going to make the radiologist workflow more straightforward and more efficient,” said Dr. Cook.

For more insights from Dr. Cook, watch the video below.

1. Chambon PJ, Wu C, Steinkamp JM, Adleberg J, Cook TS, Langlotz CP. Automated deidentification of radiology reports combining transformer and “hide in plain sight” rule-based methods. 

 2022 Nov 23;ocac219. doi: 10.1093/jamia/ocac219 . Online ahead of print.

In a recent video interview from the Radiological Society of North America (RSNA) conference, Tessa Cook, MD, PhD discussed new research on automated de-identification in radiology reports and the potential of artificial intelligence (AI) and natural language processing (NLP) to help address time-consuming challenges in the radiology workflow.

Artificial intelligence (AI) could play a key role in improving the consistency of breast density assessment, according to new research presented at the recent Radiological Society of North America (RSNA) conference.

In a retrospective, longitudinal study that examined the effectiveness of an emerging artificial intelligence (AI) model (WRDensity, Whiterabbit.ai ) for assessing breast density, researchers reviewed mammography data from over 61,000 patients who had three or more mammograms over a five-year period. They found the AI model was over 24 percent more consistent than radiologist assessment in assessing breast density via the Breast Imaging Reporting and Data System (BI-RADS).

In a recent video interview, Susan Holley, MD, a co-author of the study who presented the findings at the RSNA conference, said the AI model can help address the “innate variability” that can occur with radiologist assessment of breast density. As Dr. Holley pointed out, adjunctive AI assessment could be beneficial in risk stratification, determining the potential need for supplemental screening and documenting breast density assessment in radiology reports.

“With breast density being so important for risk assessment, supplemental screening and also now with a national breast density notification law that is looking like it’s going to actually happen in 2023, looking at variability and how we can get better at density assessment is even more timely and more critical,” noted Dr. Holley, a radiologist who is affiliated with UNC Health Care in Raleigh, N.C.

For more insights from Dr. Holley, watch the video below.

In a recent video interview, Susan Holley, MD discussed key findings from a large retrospective longitudinal study, presented at the recent Radiological Society of North America (RSNA) conference, which found that an emerging artificial intelligence (AI) model was over 24 percent more consistent than radiologist assessment of breast density. 

Can AI Improve the Consistency of Breast Density Assessment by Radiologists?

For Raymond Y. Kwong, MD, MPH, FACC, FSCMR, long scan times with cardiac magnetic resonance imaging (MRI) can have a significant downstream effect on patient access to this technology. He noted the increased demand for cardiac MRI has led to wait times of three to five weeks for outpatients and four to five days for in-patients to be scheduled for a cardiac MRI.

“These numbers are really unacceptable for patient care from our standpoint,” noted Dr. Kwong, the director of Cardiac Magnetic Resonance Imaging in the Cardiovascular Division of the Department of Medicine at Brigham and Women’s Hospital in Boston, and a professor medicine at Harvard Medical School.

With this in mind, Dr. Kwong recently started using Vista.ai (formerly HeartVista) One-Click MRI, an artificial intelligence (AI)-enabled software that helps improve image quality as well as the efficiency of cardiac MRI procedures. At the recent Radiological Society of North America (RSNA) conference, Dr. Kwong shared his six-month experience with Vista.ai’s One-Click MRI and noted a 31 percent reduction in scan times based on 1,100 consecutive cardiac MRI studies performed between April 2022 and September 2022 for patients with cardiomyopathy or structural heart disease.

When employing full use of the Vista.ai One-Click MRI software, Dr. Kwong found that 90 percent of cardiac MRI studies could be completed in less than 45 minutes in comparison to 25 percent of non-AI-aided studies.

(Editor’s note: This interview was recorded prior to HeartVista’s announcement about the company name change to Vista.ai.)

For more insights from Dr. Kwong, watch the video below.

In a recent video interview, Raymond Y. Kwong, MD, discussed his clinical experience with the Vista.ai (formerly HeartVista) One Click MRI software and recent research, presented at the Radiological Society of North America (RSNA) conference, that revealed a 31 percent decrease in cardiac MRI scan times for patients with cardiomyopathy or structural heart disease.

Can Emerging AI-Guided Software Rein in Scan Times for Cardiac MRI?

Morris Panner didn’t want to call the familiar PACS system outdated or antiquated. However, he did suggest that radiology is on the cusp of a revolution with medical image and data management that parallels the ongoing emergence of the radiologist from the reading room to being a key member of the interdisciplinary team in facilitating optimal patient outcomes.

“The PACS system is no longer a new invention. It is important and you can always improve it but what people are really looking for now is how to free that data, take that pixel information, share it, use it with other types of data and advance the liquidity and flexibility in data management. That was the revolution that came out of the SIIM (Society for Imaging Informatics in Medicine) meeting this year,” noted Panner, the president of Intelerad Medical Systems, in a recent video interview.

recent survey research showing an increasing patient acceptance of artificial intelligence (AI)

, and the more gradual “try it” phase for many radiologists as they discern the most optimal fit for AI in their current workflows. To that end, Panner also reviewed key aspects of Intelerad’s Enterprise Imaging and Informatics Suite. It is Panner’s hope that by emphasizing improved efficiency and quality in radiology workflows, the system can “free up brain space for great diagnostics.”

For more insights from Morris Panner, watch the video below.

In a video interview, Morris Panner, the president of Intelerad Medical Systems, discussed key observations from the recent Society for Imaging Informatics in Medicine (SIIM) conference, recent research about artificial intelligence (AI) adoption and emerging goals for enhancing the efficiency of radiology workflows.

AI

Latest News