Related Experiment Video
Updated: May 5, 2026
3D Ultrasound Imaging: Fast and Cost-effective Morphometry of Musculoskeletal Tissue
Published on: November 27, 2017
Lower Extremity Growth according to AI Automated Femorotibial Length Measurement on Slot-Scanning Radiographs in
John R Zech1, Laura Santos1, Steven Staffa1
1From the Department of Radiology, New York University Langone Health, 301 E 17th St, New York, NY 10010 (J.R.Z.); Departments of Radiology (L.S., D.J.) and Orthopedic Surgery (K.A.R.), Columbia University Irving Medical Center, New York, NY; and Departments of Anesthesiology (S.S., D.Z.), Surgery (S.S., D.Z.), and Radiology (A.T.), Boston Children's Hospital, Harvard Medical School, Boston, Mass.
Artificial intelligence (AI) created updated pediatric lower extremity growth standards using diverse patient data from radiographs. These AI-derived growth curves provide more accurate predictions than traditional methods.
Area of Science:
- Pediatric radiology
- Medical imaging analysis
- Artificial intelligence in medicine
Background:
- Existing pediatric lower extremity growth standards are based on outdated and limited datasets.
- Artificial intelligence (AI) offers a method to develop more current and representative growth standards.
Purpose of the Study:
- To develop an AI model for measuring pediatric lower extremity length using standing slot-scanning radiographs.
- To compare AI-derived growth curves with the conventional Anderson-Green method for accuracy.
Main Methods:
- A Mask Region-based Convolutional Neural Network was trained to segment femur and tibia on radiographs for length measurement.
- AI measurements were used to generate quantile polynomial regression growth curves.
- AI-derived curves were compared to Anderson-Green method curves for 90% growth distribution coverage.
Main Results:
- AI measurements demonstrated high accuracy with mean absolute errors of 0.25 cm (femur), 0.27 cm (tibia), and 0.33 cm (lower extremity).
- AI-derived growth curves showed significantly better coverage of the central 90% of growth in an external test set (86.7%) compared to the Anderson-Green method (73.4%).
- The study included 1874 examinations from 523 diverse pediatric patients.
Conclusions:
- AI-driven measurement of lower extremity length from radiographs can generate more accurate pediatric growth curves.
- Updated growth standards derived from diverse datasets and AI are crucial for precise pediatric growth assessment.
