Assessing Acetabular Index Angle in Infants: A Deep Learning-Based Novel Approach.

Area of Science:

• Medical Imaging
• Artificial Intelligence
• Pediatric Orthopedics

Background:

• Developmental dysplasia of the hip (DDH) is a common infant hip disorder requiring early diagnosis for effective treatment.
• Accurate diagnosis of DDH relies on expert interpretation of pelvic X-ray scans, which can be challenging without specialized training.
• Current diagnostic methods for DDH may lack objectivity and consistency.

Purpose of the Study:

• To develop and validate a computational framework for the objective detection of DDH in infant pelvic X-rays.
• To utilize a deep learning approach for precise measurement of the acetabular index angle, a key indicator of DDH.
• To create an accessible tool for medical specialists to aid in the early and accurate diagnosis of DDH.

Main Methods:

• A two-stage deep learning pipeline combining instance segmentation and keypoint detection models was employed.
• The framework analyzes infant pelvic X-ray images to identify hip abnormalities indicative of DDH.
• The system quantifies the acetabular index angle, providing objective diagnostic metrics.

Main Results:

• The deep learning model demonstrated high accuracy in measuring the acetabular angle, with an average pixel error of 2.862 ± 2.392.
• The acetabular angle measurement error was within a range of 2.402 ± 1.963° compared to ground truth annotations.
• The developed model provides an objective and unified approach to DDH diagnosis.

Conclusions:

• The proposed deep learning framework offers a reliable and objective method for detecting DDH in infants.
• Integration into a mobile application will enhance accessibility for medical specialists, reducing diagnostic burden.
• This technology has the potential to improve early DDH detection rates, leading to better patient outcomes.