Improving Accuracy in Radiology Images and Reports

Inaccurate diagnostic radiology results can lead to treatment delays, poor outcomes, higher healthcare costs, lost revenue, and operational inefficiencies. Still, more than 20 percent of radiology reports contain some sort of error.1 Estimates vary, but the average rate of diagnostic error from these reports ranges from three to five percent. That's roughly 40 million imaging diagnostic errors each year.2

To reduce these costly errors, it is imperative to improve the accuracy of radiology images and reports. That begins with identifying the barriers to accuracy.

Assessing the Problem

Why are there so many errors? Several factors come into play, and workload may top the list. Radiologists must read and interpret hundreds of images a day. One study posited that, on average, radiologists need to interpret one image every three to four seconds to keep their workday on track. This increases the probability of human error.3

Other challenges include:

  • Poor image quality
  • Lack of support for radiology staff
  • Transcription errors
  • Limited process standardization
  • Poorly written and/or confusing exam notes

Fortunately, we are entering a new era that is enabling better healthcare with transformative technology—technology that helps eliminate the causes of these challenges.

Transforming Imaging with AI

It could be argued that the need for greater clinical efficacy and efficiency has driven the enthusiasm for artificial intelligence (AI) in radiology.3 It has certainly been the result.

Radiology is ideally suited for AI applications because it is data-driven. AI can, among other things, streamline workflow, shorten reading time, lead to earlier detection of disease, and improve diagnostic accuracy.4 When it comes to ultrasound, AI has the potential to improve workflow efficiency by enhancing the quality of radiology images, providing diagnostic support, and offering scan guidance, for example.5

By automating these activities, AI removes some of the pressure from radiologists, allowing them to focus on quality, not quantity. These AI-driven efficiencies yield other benefits, including an automated extraction of quantitative information, something often considered too time-consuming to perform routinely.6

All this may help radiologists function more efficiently and accurately, especially in terms of ultrasound. New AI tools and systems aim to dramatically improve the data throughput processing power of previous systems, enabling radiologists to segment lesions and identify vessels with ease. The objective? Faster, more accurate diagnoses and better outcomes for patients.

Enhancing Performance with Collaboration

If AI is about machine learning, virtual connectivity is about human learning. Enabling training and support from colleagues—in real time—can transform the way ultrasound teams learn and perform. Such collaboration can help improve workflow efficiency and clinical accuracy while driving standardization. For instance, a new sonographer could discretely ask for guidance before, during, or after an exam, avoiding costly errors and learning from the experience of others.

Such tools also allow radiologists to quickly connect with each other or sonographers if they need additional information or clarification about system tools on radiology images. This, too, helps ensure consistency and standardization within and across teams.

Streamlining Reporting with Tech

Another key moment in the radiology workflow where accuracy could be boosted is during the transfer of exam data into the report. The final reports for ultrasound exams include many measurements and details. Frequently, radiologists dictate those measurements from a handwritten sonographer worksheet scanned into PACS (the picture archiving and communication system). That's not only time-consuming, but it also leaves considerable room for error.

Advanced technology helps eliminate the potential for errors and streamlines the workflow by transferring data directly into the dictation system. When ultrasound data transmits directly from the scanner and populates in ultrasound reports, radiologists don't have to worry about interpreting a handwritten worksheet or double-checking images to make sure a number is correct. Radiologists can spend more time on what they do best: interpreting images.

Paving the Way for Improved Accuracy

By targeting and solving individual problems, radiology teams can be better positioned to reduce errors in ultrasound images and reports. Specifically, integrating AI and other digital tools in the imaging process can improve patient care, enhance efficiency, and advance the profession. Such solutions are no longer "nice to have"—they represent a much-needed path forward to better patient outcomes.


A Glimpse into the Digital Future of Ultrasound: Hear from Karley Yoder on the state of the Ultrasound Digital & AI market complemented by a fireside chat with Dr. Jason Wiesner at Sutter Health and the complex challenges they are tackling today to support improved care now and in the future.


1. Frequency and Spectrum of Errors in Final Radiology Reports Generated With Automatic Speech Recognition Technology; Quint, Leslie E. et al.; Journal of the American College of Radiology, Volume 5, Issue 12, 1196 – 1199.  

2. Itri JN, Tappouni RR, McEachern RO, et al. Fundamentals of diagnostic error in imaging. RadioGraphics. October 2018;38(6). doi: 10.1148/rg.2018180021.

3. Hosny A, Parmar C, Quackenbush J, et al. Artificial intelligence in radiology. Nature Reviews Cancer. May 2018;18:500-510. doi: 10.1038/s41568-018-0016-5.

4. van Leeuwen KG, de Rooij M, Schalekamp S, et al. How does artificial intelligence in radiology improve efficiency and health outcomes? Pediatric Radiology. June 2021;52:2087-2093. doi: 10.1007/s00247-021-05114-8.

5. Park SH. Artificial intelligence for ultrasonography: Unique opportunities and unique challenges. Ultrasonography. January 2021;40(1):3-6. doi: 10.14366/usg.20078.

6. Guilford-Blake R. Wait. Will AI replace radiologists after all? Radiology Business. Published February 18, 2020. Accessed March 2, 2023.