By Elodie Vaz | Published on March 19, 2026 | 3 min readElbow
injuries are a common reason for consultation in pediatric emergency
departments. However, their radiological diagnosis remains challenging. In
children, the skeleton is still developing and includes numerous growth plates
that are invisible or poorly visible on X-rays. In addition, some fractures can
be extremely subtle.
In this
context, interpreting elbow radiographs is a real clinical challenge, even for
experienced practitioners. A missed fracture can lead to inappropriate
management and, ultimately, functional complications. Faced with this
diagnostic difficulty, teams from the pediatric emergency department and the
radiology department at Nantes University Hospital explored an innovative
approach: integrating an artificial intelligence (AI) algorithm as a
decision-support tool.
Assessing
the contribution of AI in fracture detection
The
objective of the study conducted at Nantes University Hospital was to measure
the impact of a deep learning algorithm on clinicians’ diagnostic performance
in detecting elbow fractures in children.
The
researchers retrospectively analyzed radiographs from 755 children aged 0 to 15
years who were treated in the pediatric emergency department for elbow trauma
between January 2019 and April 2020. The goal was to determine whether the use
of an AI system could improve diagnostic sensitivity and reduce the risk of
missed fractures in an emergency setting.
A
comparative methodology
To
establish a reference diagnosis, two independent experts reviewed all
radiological examinations without knowledge of the algorithm’s results.
Diagnoses
made by emergency physicians were then compared across three configurations:
without technological assistance, with theoretical assistance from the AI
algorithm, and with the algorithm used alone. This comparative approach aimed
to precisely quantify the contribution of artificial intelligence to clinical
practice.
Significantly
improved diagnostic performance
The
results, published last January in the European Journal of Radiology,
show a significant impact of AI on fracture detection. Algorithm-assisted
interpretation achieved a diagnostic sensitivity of 99%, reflecting
near-complete detection of fractures when clinicians were supported by the
tool.
Moreover,
the study demonstrated a gain in sensitivity of over 20% for emergency
physicians when assisted by AI. This improvement directly translates into a
reduced risk of missed fractures, a major issue in pediatric trauma care.
According
to Dr. Fleur Lorton, pediatrician in the pediatric emergency department at
Nantes University Hospital and author of the study, AI should be viewed as a
complement to medical expertise:
“Interpreting
elbow radiographs in children is a particularly demanding task in pediatric
trauma care. Artificial intelligence does not replace the physician; it acts as
a second reader—a co-pilot that enhances our analysis. Through this
human–machine collaboration, we reduce the risk of error and improve the
quality of care for young patients,” she explained in a press release from
Nantes University Hospital published on March 9.
Toward
broader integration of AI in pediatric emergency medicine
This work,
resulting from collaboration between pediatric emergency, radiology, and the
Women–Children–Adolescents Clinical Investigation Center at Nantes University
Hospital, highlights the potential of AI tools in medical imaging.
Following
this study, an AI-based software dedicated to trauma care is now used in
routine practice in the hospital’s pediatric emergency department. It can
automatically identify and localize various abnormalities on limb radiographs,
including fractures, dislocations, and joint effusions.
A new study
is also underway to evaluate an algorithm applied to chest radiographs, with
the aim of extending these tools to acute respiratory conditions, which are
particularly common in pediatrics.
AI as a
“second clinical reader”
The Nantes
study confirms the potential of artificial intelligence as a decision-support
tool in emergency settings. By enhancing clinicians’ diagnostic sensitivity,
these systems can help secure radiological interpretation and improve the
management of young patients.
Read next: Pap smear & AI: the new era of cervical screening
About the Author – Elodie Vaz
Health journalist, CFPJ graduate (2023).
Élodie explores the marks diseases leave on bodies and, more broadly, on human life. A registered nurse since 2010, she spent twelve years at patients’ bedsides before exchanging her stethoscope for a notebook. She now investigates the links between environment and health, convinced that the vitality of life cannot be reduced to that of humans alone.
