Technology Report:

Artificial Intelligence

Technology Report: Artificial Intelligence in Radiology 2021

This is an overview of trends and technologies in radiology artificial intelligence (AI) applications in 2021. Views were shared by 11 radiologists using AI and industry leaders, which include:

• Randy Hicks, M.D., MBA, radiologist and CEO of Reginal Medical Imaging (RMI), and an iCAD Profound AI user.

• Prof. Dr. Thomas Frauenfelder, University of Zurich, Institute for Diagnostic and Interventional Radiology, and Riverain AI user.

• Amy Patel, M.D., medical director of Liberty Hospital Women’s Imaging, assistant professor of radiology at UMKC, and user of Kios AI for breast ultrasound.

• Sham Sokka, Ph.D., vice president and head of innovation, precision diagnosis, Philips Healthcare.

• Ivo Dreisser, Siemens Healthineers, global marketing manager for the AI Rad Companion.

• Bill Lacey, vice president of medical informatics, Fujifilm Medical Systems USA.

• Karley Yoder, vice president and general manager, artificial intelligence, GE Healthcare.

• Georges Espada, head of Agfa Healthcare digital and computed radiography business unit.

• Pooja Rao, head of research and development and co-founder of Qure.ai.

• Jill Hamman, world-wide marketing manager at Carestream Health.

• Sebastian Nickel, Siemens Healthineers, global product manager for the AI Pathway Companion.

There has been a change in attitudes about AI on the expo floor at the Radiological Society of North America (RSNA) over the last two years. AI conversations were originally 101 level and discussed how AI technology could be trained to sort photos of dogs and cats. However, in 2020, with numerous FDA approvals for various AI applications, the conversations at RSNA, and industry wide, have shifted to that of accepting the validity of AI. Radiologists now want to discuss how a specific AI algorithm is going to help them save time, make more accurate diagnoses and make them more efficient.

With a higher level of maturity in AI and the technology seeing wider adoption, radiologists using it say AI gives them additional confidence in their diagnoses, and can even help readers who may not be deep experts in the exam type they are being asked to read.

With a myriad of new AI apps gaining regulatory approval from scores of imaging vendors, the biggest challenge for getting this technology into hospitals is an easy to integrate format. This has led to several vendors creating AI app stores. These allow AI apps to integrate easily into radiology workflows because the apps are already integrated as third-party software into a larger radiology vendors' IT platform.

There are now hundreds of AI applications that do a wide variety of analysis, from data analytics, image reconstruction, disease and anatomy identification, automating measurements and advanced visualization. The AI applications can be divided into 2 basic types — AI to improve workflow, and AI for clinical decision support, such as diagnostic aids.

On the workflow side, several vendors are leveraging AI to pull together all of a patients' information, prior exams and reports in one location and to digest the information so it is easier for the radiologist to consume. Often the AI pulls only data and priors that relate to a specific question being asked, based on the imaging protocol used for the exam. One example of this is the Siemens Healthineers AI Clinical Pathway and Siemens AI integrations with PACS to automate measurements and advanced visualization.

AI is also helping simplify complex tasks and help reduce the reading time on involved exams. One example of this is in 3-D breast tomosythesis with hundreds of images, which is rapidly replacing 2-D mammography, which only produces 4 images. Another example is automated image reconstruction algorithms to significantly reduce manual work. AI also is now being integrated directly into several vendors' imaging systems to speed workflow and improve image quality.

Vendors say AI is here to stay. They explain the future of AI will be automation to help improve image quality, simplify manual processes, improved diagnostic quality, new ways to analyze data, and workflow aids that operate in the background as part of a growing number of software solutions.

Several vendors at RSNA 2020 noted that AI's biggest impact in the coming years will be its ability to augment and speed the workflow for the small number of radiologists compared to the quickly growing elder patient populations worldwide. There also are applications in rural and developing countries were there are very low numbers of physicians or specialists.

Sidebar Videos

Related Content

Sarcopenia (Myosteatosis) at Screening CT Colonoscopy in 79-Year-Old Woman With Subsequent Hip Fracture

Sarcopenia (Myosteatosis) at Screening CT Colonoscopy in 79-Year-Old Woman With Subsequent Hip Fracture: A & B. Image at L1 vertebral level without (A) and with (B) overlay of automated skeletal muscle segmentation (B, red). Intramuscular fat is present within paraspinal muscles (A, yellow circle). Mean muscle attenuation is similar for manually placed ROI (1.8 HU) and automated tool (3.9 HU), being markedly decreased for both approaches. Muscle cross-sectional area is relatively preserved when intramuscular fat is included.

C & D. CT images at L3 level without (C) and with (D) muscle segmentation demonstrate similar findings.

E. Hip radiograph demonstrates left intertrochanteric femoral fracture that occurred when patient fell 2 years later. Patient died 4 years after falling.

News | August 19, 2021

Deep Learning Tool for Sarcopenia on CT: L1 Versus L3 Vertebral Level

qER-Quant will help enable faster diagnosis and help track patient progression over time

News | August 17, 2021

Qure.ai Receives Second FDA Clearance for Brain CT Scan AI

August 17, 2021 — Qure.ai has announced US FDA 510(k) clearance for

#coronavirus #COVID19 #pandemic #RSNA21 will require all attendees and exhibitors to show proof of #vaccination and wear a #mask

Getty Images

News | August 12, 2021

RSNA21 Will Require Proof of Vaccination, Masks

August 12, 2021 — RSNA 2021 will require proof of CO...

Figure shows coarse attention maps generated using GradCAM for correctly classified high-grade glioma (HGG), low-grade glioma (LGG), brain metastases (METS), meningioma (MEN), acoustic neuroma (AN), and pituitary adenoma (PA). For each pair, the postcontrast T1-weighted scan, and the GradCAM attention map (overlaid on scan) have been shown. In GradCAM maps, warmer and colder colors represent high and low contribution of pixels toward a correct prediction, respectively. Image courtesy of the Radiological Society of North America

News | August 12, 2021

Deep Learning Model Classifies Brain Tumors with Single MRI Scan

August 12, 2021 — A team of researchers at Washington Universit...

artificial intelligence-based computer-aided detection (AI-CAD) can be a practical addition for lowering false-positive findings when performing post-breast conserving therapy (BCT) surveillance mammography

AI-CAD marked axillary lymph node and region in right upper outer quadrant (arrows and thin line outlining both sites) and assigned an abnormality score of 28%.

News | August 11, 2021

AI DBT Impact on Mammography Post-breast Therapy