Journal highlights

The following are highlights from the current issues of RSNA’s peer-reviewed journals.

Radiology Logo

Chatbots and Large Language Models in Radiology

Chatbots, originally developed in 1966, are applications designed to mimic human conversation. Advances in natural language processing led to the release of virtual assistants like Siri and Google Assistant in the early 2010s. These assistants helped users conduct simple tasks but were limited in their understanding and generation of language.

OpenAI’s ChatGPT, launched in late 2022, initiated the latest wave of AI chatbots that could change the way humans interact with technology and revolutionize many industries. The performance and flexibility of these chatbots can be attributed to the rapidly advancing large language models (LLMs) that underlie them.

In an article published in Radiology, Rajesh Bhayana, MD, Toronto General Hospital, reviews the potential uses and limitations of LLMs and mitigation strategies. He also examines existing LLM-based applications that can enhance efficiency in supervised settings.

Dr. Bhayana noted that LLMs offer many clinical and research applications in radiology including, simultaneously interpreting text and images to generate reports and overall improving the efficiency, cost effectiveness and quality of care in supervised practice settings.

However, the models have limitations that include generating inaccurate responses, knowledge cutoff dates, poor complex reasoning and a tendency to perpetuate bias and generate different responses to the same prompt.

“LLMs and the AI chatbots built on them have rapidly evolved. Since LLMs have limitations, tools optimized to overcome these limitations should be investigated and validated for use in radiology. With the increasing use of AI chatbots and LLMs across industries, radiologists should be leaders in shaping how these tools are leveraged in our specialty,” Dr. Bhayana concludes.

Read the full article, “Chatbots and Large Language Models in Radiology: A Practical Primer for Clinical and Research Applications.” Follow the Radiology editor on X @RadiologyEditor.

Fig 3 from Radiology study by Bhayana et al examining the use of chatbots and large language models in radiology

Diagram shows potential clinical applications of large language models (LLMs) in radiology, from guiding the decision to perform imaging to helping clinicians and patients interpret reports. OpenAI’s DALL-E 3 (https://openai.com/dall-e-3) was used to generate the brain neural network icon in this figure. Hx = History.

https://doi.org/10.1148/radiol.232756 ©RSNA 2024

Radiograpics

Using CT and MRI Enterography to Diagnose Celiac Disease

Celiac disease is a common autoimmune inflammatory disease of the small bowel that causes mucosal intestinal lesions. It is triggered by the ingestion of gluten in genetically predisposed individuals.

Gluten contains gliadin, a component found mostly in wheat, barley and rye. Gluten ingestion can lead to gastrointestinal malabsorption with symptoms such as diarrhea, constipation, abdominal pain and distention. While recognition of celiac disease has increased, it remains challenging to diagnose.

In an article published in RadioGraphics, Andrea Penizzotto, MD, Hospital Italiano de Buenos Aires, in Argentina, and colleagues review the diagnosis of celiac disease and its complications. Their work includes a particular focus on findings from CT and MR enterography, noninvasive studies used to evaluate small bowel pathologic conditions.

The authors recognize celiac disease can be difficult to diagnose, as many symptoms are nonspecific. They note that CT and MR enterography present certain imaging features that can aid diagnosis. These include an alteration of the small bowel fold pattern, flattening of the mucosa, thickening of the wall and folds in the small bowel, and intestinal dilatation.

“Patients with celiac disease present with nonspecific symptoms and a broad spectrum of imaging findings. However, the radiologist may help in the diagnosis of celiac disease by carefully examining the small bowel and extraintestinal findings. The presence of ulcerative jejunoileitis, malignancy, cavitated mesenteric lymph nodes, or hyposplenism is an indicator of advanced disease,” the authors conclude.

Read the full article, “CT and MR Enterography in the Evaluation of Celiac Disease.” This article will also be available for CME. Follow the RadioGraphics editor on X @RadG_Editor.

Figures 18 A to C from RadioGraphics study by Penizzotto et al showing Type II refractory celiac disease in a 68 year-old man with abdominal pain

Type II refractory celiac disease in a 68-year-old man with abdominal pain. (A, B) Axial (A) and coronal (B) CT enterographic images with intravenous contrast agent administration show asymmetric and segmental thickening of the jejunal wall, with soft tissues extending into the perienteric tissues (arrows). (C) Axial 18F FDG PET/CT fusion image shows increased uptake in the primary tumor (arrow) and a left internal iliac adenopathy (arrowhead). Jejunal resection was performed, and results demonstrated T-cell lymphoma.

https://doi.org/10.1148/rg.230122 ©RSNA 2024

 

Assessing Fat Density Around Abdominal Aorta in Aneurysms

Consuming too many calories can lead to changes in adipose tissue by increasing the number and size of fat cells, leading to cell dysfunction and apoptosis. The resulting inflammation and development of fibrous tissue creates a chronic state of low-grade inflammation in the body. Analyzing the density of fat tissue around coronary arteries can help predict heart attacks and cardiac-related mortality.

In a study published in Radiology: Cardiothoracic Imaging, Samuel Debono, MD, University of Edinburgh Centre for Cardiovascular Science, and colleagues sought to assess periaortic adipose tissue attenuation at CT imaging in different abdominal aortic aneurysm disease states.

They conducted a retrospective observational study of CT angiography of 70 patients with abdominal aortic aneurysms (30 symptomatic and 40 asymptomatic) and 18 without. They measured adipose tissue attenuation in different segments of the abdominal aorta and in subcutaneous and visceral fat and compared adipose tissue attenuation values across all patient groups.

Periaortic adipose tissue CT attenuation was not increased in stable abdominal aortic aneurysm disease. There was a generalized increase in attenuation in patients with symptomatic disease, likely reflecting the systemic consequences of acute rupture. “This suggests that periaortic adipose tissue attenuation may be of limited clinical value as a biomarker or risk stratification tool in abdominal aortic aneurysm disease. Further exploration of the implications and mechanisms underlying the generalized changes in adipose tissue attenuation in patients with symptomatic aneurysm disease is warranted,” the researchers conclude.

Read the full article, “CT Attenuation of Periaortic Adipose Tissue in Abdominal Aortic Aneurysms. ” Follow the Radiology: Cardiothoracic Imaging editor on X @RadiologyCTI.

Submit Your Work to RSNA Journals

RSNA publishes six peer-reviewed journals, keeping readers up to date on the latest advances in radiology. Give your work the visibility it deserves—submit your manuscript for consideration in one of the following:

  • Radiology – RSNA’s premier publication featuring the most current, clinically relevant and highest quality research in the field
  • RadioGraphics – RSNA’s primary education journal featuring articles across radiologic subspecialties, including imaging physics, informatics and diagnostic imaging. Each issue includes opportunities for CME
  • Radiology: Artificial Intelligence – Special focus journal featuring topics including the impact of AI on diagnostics and patient care, AI’s role in radiology education and the ethical implications presented by AI in medicine.
  • Radiology: Cardiothoracic Imaging –Special focus journal centered on advances in medical imaging that drive cardiothoracic medicine
  • Radiology: Imaging Cancer – Special focus journal featuring cancer screening, differential diagnosis and treatment planning and featuring an interdisciplinary perspective on cancer screening
  • Radiology Advances – RSNA’s newest and first exclusively open-access journal publishing primarily original multidisciplinary research articles with a focus on emerging topics, cross-cutting and innovative research.

Learn more about how to submit your work to RSNA journals.