Automated neonatal hip ultrasound system for diagnosing developmental dysplasia of hips using assistive AI

Abstract Children with developmental dysplasia of the hip (DDH) joint often do not exhibit clinical symptoms, and diagnosis in some cases relies solely on hip joint examination. There are several methods to examine the hip joint, with ultrasound being the simplest and most convenient diagnostic method. The Graff algorithm is commonly used to diagnose hip dysplasia by measuring angles at different landmarks in the hip region. This study utilized the SONON 300L device developed by Hilserion to acquire ultrasound images of the hip joints in 18 infants and children. Following classification, the standard planes were aligned with the Graf’s landmarks using an artificial intelligence (AI) model, and the landmark regions were subsequently segmented. Significant findings were observed in the ultrasound images, with a maximum area-under-the-curve of 0.864 and a Dice coefficient of 0.660. The Graf algorithm was employed to develop an automated model for diagnosing DDH, resulting in a DDH-assisted AI model with an average Graf angle error rate of 0.21 compared to expert diagnostics