Accurate bone age evaluation in pediatric patients is crucial for assessing skeletal maturation, but traditional methods looking at hand bones can have limitations in patients going through puberty.
Choi
Research presented at a Tuesday session proposed a more precise and feasible elbow bone age classification and uses a deep-learning AI model for bone evaluation.
Accurate bone age assessment is important for various clinical conditions as it represents skeletal maturity, which can differ from chronological age. In puberty, children undergo remarkable physical changes and growth spurts.
“Our research introduces a newly revised elbow bone age evaluation method using the olecranon apophysis, along with an AI model for puberty. Since hand bone age shows limitations in puberty, elbow bone age can be a useful alternative,” said Gayoung Choi, MD, PhD, a radiologist at Korea University in Seoul. “We believe that the findings of this research are of significant importance to the pubertal bone age evaluation and will contribute valuable insights into the accurate and practical elbow bone age evaluation for puberty, along with rapidly developing AI-assisted bone age assessment.”
Simpler And Cost-Effective For Patients
The study included 3,508 lateral elbow radiographs of patients under the age of 18 that were reviewed and classified based on the morphological changes of the olecranon, or the bony prominence of the elbow.
These were then compared with previously established elbow bone age methods (Sauvegrain and Dimeglio) and hand and wrist bone age methods and charts (Greulich and Pyle method and the Korean standard bone age chart). The deep-learning-based model for olecranon bone age classification was trained, validated, tested and further improved through external validation. Two pediatric radiologists then provided interobserver review.
Overall, the deep-learning based model showed a 96% accuracy and a 98% specificity. There was excellent reliability with both Sauvegrain and Dimeglio methods for girls and boys (ICC 0.97 and 0.98 in girls, 0.96 and 0.98 in boys each), and good reliability with hand and wrist bone age by both GP method (0.78 in girls, 0.84 in boys) and KS chart (0.87 in girls, 0.89 in boys).
The accurate diagnosis of the onset of puberty and its progression is crucial for various pediatric conditions related to skeletal maturation, Dr. Choi commented.
“This new method using olecranon apophysis showed excellent reliability with existing bone age evaluation methods and almost perfect interobserver agreement between two specialized pediatric radiologists, while requiring only a single lateral elbow radiograph,” Dr. Choi said.
Although elbow age has been less widely used than hand bone age, Dr. Choi said the hope is that this revised categorization of the olecranon ossification process and AI model could help with accurate pubertal bone age evaluation and inspire further exploration.
Relying on an elbow radiograph rather than the more in-depth procedures for hand bone analysis could also make the process simpler, less expensive and more cost effective for patients.
“Bone age is an important pediatric biological age that represents skeletal maturity and may differ from chronological age,” Dr. Choi said. “Accurate assessment of bone maturity is important for diagnosis and planning treatment for various conditions. This novel approach shows excellent reliability and can improve the process for our patients.”
Access the presentation, “A Simple and Practical Elbow Bone Age Evaluation Using Olecranon Apophysis in Puberty and Development of a Deep-Learning Model,” (T1-SSPD03-3) on demand at Meeting.RSNA.org .
