RSNA 2019 Artificial Intelligence Preview

Road to RSNA 2019: Artificial Intelligence Preview

By Erik L. Ridley, AuntMinnie staff writer
November 4, 2019

Welcome to the first installment of this year's Road to RSNA preview of the RSNA 2019 meeting in Chicago. For the 11th year in a row, we're providing a modality-by-modality overview of select scientific presentations to serve as your guide to events at McCormick Place.

Our journey along the Road to RSNA begins with our preview of artificial intelligence (AI), a topic that's once again primed to take center stage at the meeting. The widespread initial concern that AI would soon replace radiologists now seems like a distant memory, as AI research is largely focused these days on augmenting radiologists and improving the practice of radiology and, consequently, patient care.

Algorithms that serve as a triage tool -- prescreening certain types of imaging studies and alerting radiologists to the probability of a critical finding -- remain a popular area of interest. For example, in one presentation at RSNA 2019, researchers will report that deep-learning software for detecting intracranial hemorrhage can speed up report turnaround times and reduce the length of hospital stays for patients.

Research has also increased on applying AI with radiomics to assess risk, as well as monitor and predict response to treatment. In one study, researchers found that an AI algorithm can estimate future healthcare expenses for a patient based on analysis of a chest radiograph.

Importantly, there's been a flurry of activity in applying AI algorithms during the image acquisition process, showing potential for sharply lowering radiation dose and image scanning time. A number of presentations will highlight how using AI helps improve diagnostic performance for radiologists.

A type of AI algorithm called generative adversarial networks (GANs) is showing promise for a variety of applications, including producing synthetic images and enabling models to be trained with normal exams to detect anomalies. On a modality basis, mammography and radiography -- particularly chest x-rays -- remain prominent, although AI is clearly being investigated for use with all modalities.

See below for previews of these and other AI-related scientific talks at this year's RSNA meeting. Of course, these are just a sample of the content on offer; hundreds of other scientific sessions, scientific posters, refresher courses, and educational exhibits on AI topics also await attendees. For more information on those presentations and other abstracts in this year's scientific and educational program, view the RSNA 2019 meeting program.

Also, be sure to visit the AI Showcase -- formerly known as the Machine Learning Showcase. The RSNA has moved this year's event to a larger space on North Hall Level 2 to provide a central hub of AI activities for attendees.

The AI Showcase will feature well over 100 companies, a significant increase compared with last year. These firms will showcase their AI wares in over 40,000 sq ft of exhibit space in 39 20-minute vendor presentations throughout the week at the AI Showcase Theater and in the new Hands-on Classroom.

In addition, the AI Showcase will host the RSNA Deep Learning Classroom, which will offer 14 courses for practicing radiologists. What's more, a special RSNA exhibit will demonstrate how AI and other decision-support tools can be incorporated and be effective in the radiology workflow.

Scientific and Educational Presentations
Deep learning improves DBT's efficiency
Sunday, December 1 | 10:45 a.m.-10:55 a.m. | SSA01-01 | Room S406A
In this presentation, researchers will describe how using a deep-learning algorithm with digital breast tomosynthesis (DBT) can reduce radiologists' interpretation times.
Deep learning detects bone metastases on PET/CT
Sunday, December 1 | 10:45 a.m.-10:55 a.m. | SSA16-01 | Room S505AB
A deep-learning system can yield promising results for identifying and annotating bone metastases on PET/CT image data, according to this talk by researchers from Japan.
AI can opportunistically screen for vertebral fractures
Sunday, December 1 | 11:05 a.m.-11:15 a.m. | SSA12-03 | Room E450A
In this talk, researchers will present their automated method to provide opportunistic screening for vertebral fractures on CT exams.
AI can be used alone for low-malignancy mammograms
Sunday, December 1 | 11:15 a.m.-11:25 a.m. | SSA01-04 | Room S406A
An artificial intelligence (AI) algorithm can be used for standalone interpretation of mammograms that have a low probability of being malignant, researchers from the University of Southern California will report in this presentation.
Algorithm finds hip, pelvic fractures on x-rays
Sunday, December 1 | 11:15 a.m.-11:25 a.m. | SSA12-04 | Room E450A
An artificial intelligence (AI) algorithm was able to detect both hip and pelvic fractures on pelvic radiographs in a study by researchers from Taiwan.
AI-based FFR-CT adds value in triple-rule-out angio
Sunday, December 1 | 11:35 a.m.-11:45 a.m. | SSA03-06 | Room S105AB
Artificial intelligence (AI)-based fractional flow reserve (FFR-CT) analysis can add diagnostic and prognostic value in patients receiving triple-rule-out coronary CT angiography, according to a study to be presented in this talk.
AI for DBT improves breast cancer detection
Sunday, December 1 | 11:45 a.m.-11:55 a.m. | SSA01-07 | Room S406A
When radiologists use artificial intelligence (AI) with digital breast tomosynthesis (DBT), their cancer detection performance and efficiency improve, according to data to be presented Sunday morning.
Radiomics can assess metastatic liver disease treatment
Sunday, December 1 | 11:55 a.m.-12:05 p.m. | SSA08-08 | Room S104A
Radiomics and machine learning can accurately assess treatment response for liver metastases in breast cancer patients, according to this scientific presentation.
AI algorithm detects, tags, and segments lesions on CT
Monday, December 2 | 10:30 a.m.-10:40 a.m. | SSC08-01 | Room E450A
In this talk, researchers will share details about a deep-learning algorithm for detecting, tagging, and segmenting various lesions on CT images.
GANs ease 'big data' problem in training AI algorithms
Monday, December 2 | 11:40 a.m.-11:50 a.m. | SSC04-08 | Room S102CD
A type of AI technology called generative adversarial networks (GANs) that was trained using normal brain CT scans is able to detect various intracranial diseases, according to a study by researchers from South Korea.
AI can speed up whole-body diffusion-weighted MRI
Monday, December 2 | 11:40 a.m.-11:50 a.m. | SSC08-08 | Room E450A
An artificial intelligence (AI) algorithm can sharply reduce the scanning time required for whole-body diffusion-weighted MRI, according to researchers from the U.K.
AI predicts future healthcare costs from chest x-rays
Monday, December 2 | 11:50 a.m.-12:00 p.m. | SSC08-09 | Room E450A
In this talk, a group from California will reveal how an artificial intelligence (AI) algorithm can predict future healthcare expenses for patients just by analyzing their chest radiograph.
Deep learning can find osteoporosis on chest x-rays
Monday, December 2 | 11:50 a.m.-12:00 p.m. | SSC09-09 | Room E450B
A deep-learning algorithm can predict osteopenia and osteoporosis on chest radiographs, according to a team of researchers from Maryland.
ICH detection software reduces report turnaround times
Monday, December 2 | 3:10 p.m.-3:20 p.m. | SSE14-02 | Room S406B
Deep-learning software for detecting intracranial hemorrhage (ICH) on CT scans can deliver faster radiology report turnaround times and reduce patient hospital stays, according to this talk.
AI finds missed lung nodules on whole-body CT
Monday, December 2 | 3:40 p.m.-3:50 p.m. | SSE14-05 | Room S406B
In this scientific presentation, a team of researchers from Taiwan will report on how an artificial intelligence (AI) algorithm served as a helpful second reader for whole-body CT exams.
Deep-learning models could be fooled by fake images
Tuesday, December 3 | 9:05 a.m.-9:15 a.m. | RC315-04 | Arie Crown Theater
Fake images produced by generative adversarial networks (GANs) have the potential to fool deep-learning algorithms and sometimes even radiologists, according to this talk.
AI doesn't hedge when reporting on chest x-rays
Tuesday, December 3 | 11:40 a.m.-11:50 a.m. | SSG06-08 | Room S406A
In this presentation, researchers from India will describe how a deep-learning algorithm shows potential for producing accurate reports of chest radiographs.
AI algorithm could enhance CT lung cancer screening
Tuesday, December 3 | 3:50 p.m.-4:00 p.m. | SSJ05-06 | Room S102CD
An artificial intelligence (AI) algorithm could help radiologists improve their diagnostic accuracy on CT lung cancer screening exams, according to this presentation.
1st-year radiology residents beat AI for detecting pneumothorax
Wednesday, December 4 | 3:00 p.m.-3:10 p.m. | SSM07-01 | Room N228
Although it was much faster, an artificial intelligence (AI) algorithm couldn't beat the performance of first-year radiology residents for detecting pneumothorax in a study by researchers from Johns Hopkins University.
AI enhances multiparametric MRI reads
Wednesday, December 4 | 3:20 p.m.-3:30 p.m. | SSM12-03 | Room N229
Artificial intelligence (AI) can improve the sensitivity of radiologists for detecting intraprostatic lesions on multiparametric MRI, according to a multinational team of researchers.
AI can help thwart cyberattacks on imaging devices
Wednesday, December 4 | 3:20 p.m.-3:30 p.m. | SSM15-03 | Room E353B
In this presentation, researchers from Israel will describe how artificial intelligence (AI) technology can help protect medical imaging devices from cyberattacks.
Patients need to be educated on AI use in radiology
Wednesday, December 4 | 3:40 p.m.-3:50 p.m. | SSM13-05 | Room S404CD
In this talk, researchers from Canada will report on patient perspectives regarding the use of artificial intelligence (AI) technology in radiology.
GAN model predicts interstitial lung disease survival
Wednesday, December 4 | 3:40 p.m.-3:50 p.m. | SSM14-05 | Room E353C
In this scientific session, researchers from Massachusetts will describe how generative adversarial networks (GANs) can be utilized to predict survival in patients with idiopathic pulmonary fibrosis.
AI converts microdose CT to virtual high-dose images
Thursday, December 5 | 10:40 a.m.-10:50 a.m. | SSQ19-02 | Room E353B
A deep-learning technique can create virtual high-dose CT images from microdose images, helping reduce radiation dose delivered to patients, according to this Thursday presentation.
Machine-learning model predicts interpretation delays
Thursday, December 5 | 11:00 a.m.-11:10 a.m. | SSQ11-04 | Room N229
A machine-learning algorithm can predict delays in performing and interpreting on-call radiology exams, a team of California researchers will report in this presentation.
Deep-learning model could triage renal ultrasound exams
Thursday, December 5 | 11:10 a.m.-11:20 a.m. | SSQ10-05 | Room E352
A deep-learning algorithm can detect abnormalities on renal ultrasound exams, potentially enabling triage of these cases for radiologists, according to this scientific presentation.
AI classifies IDH mutation status in brain tumors
Thursday, December 5 | 11:30 a.m.-11:40 a.m. | SSQ15-07 | Room S404AB
An artificial intelligence (AI) algorithm was able to classify the mutation status of isocitrate dehydrogenase (IDH) in brain tumors in a study by researchers from Texas.