Bone growth encompasses alterations in bone size, shape, and mineral density. This occurs through the activity of primary and secondary centres of ossification, the bone formed from the first centre is known as the diaphysis, and from the second, the epiphysis, respectively. In these centres, cartilage gradually transforms into bone tissue. This progression continues as long as cartilage remains present in the growth plate, also known as the epiphyseal plate. Upon completion of bone development, the epiphyseal plate undergoes ossification, indicating fusion between the diaphysis and epiphysis [1]. Bone age serves as a marker of bone maturity making its assessment common in paediatric radiology. It aids in evaluating growth, maturity, and diagnosing and managing various paediatric disorders, including endocrinological, orthodontic, and orthopaedical conditions. Accurate assessment relies on understanding the shape and maturity level of primary and secondary ossification centres and their fusion times [2, 3]. The two primary applications of skeletal age assessment are the identification of growth disorders and the estimation of eventual adult height. From a legal standpoint, bone age assessment could play a role in determining whether an individual is a minor when official documents are unavailable. However, according to the European Society of Paediatric Radiology (ESPR), evaluating the bone age of the hand and wrist alone to determine chronological age is not recommended because it is not possible to overcome the large biological variation or the statistical problems associated with endpoint maturation of the wrist [4].

Over the past decades, various methods have been utilized including the Greulich–Pyle (GP), the Gilsanz–Ratibin, and the Tanner–Whitehouse (TW) methods. The GP and the Gilsanz–Ratibin methods are atlas-based, comparing the patient's radiograph to standard atlas radiographs and assigning the nearest bone age [1, 5, 6]. Conversely, TW employs a scoring method, staging specific radiographic regions of interest (ROI) of the radius, ulna, and short bones, to derive a final score converted into bone age [7]. Greulich and Pyle's Radiographic Atlas of Skeletal Development of the Hand and Wrist (G&P) presents left-hand radiographs chosen as sex-specific developmental benchmarks across various ages. The atlas includes tables of mean skeletal ages and standard deviations (SD), categorized by chronological ages and sex, facilitating assessments of skeletal maturity in children [6]. Greulich and Pyle curated representative radiographs to correspond with each age group in the atlas. By comparing these standards with radiographs from hundreds of typically developing children of similar ages, they calculated standard deviations for each age group. Despite GP's creation using radiographs from the forties and fifties, it continues to be widely used in clinical practice, albeit requiring manual processing, and is applicable to multi-ethnic populations in developed countries [8, 9]. The manual approach of the G&P method involves reviewing images and text, consulting data tables, and performing basic calculations to assess skeletal age against chronological age. This manual process can slow down diagnostic workflows and increase the risk of both observer and mathematical errors [8].

BoneXpert® (Visiana, Holte, Denmark) is a fully automated system that operates without the need for manual verification by an expert, introduced in 2009. This software adopts a more nuanced approach, examining radiographs of the left hand and wrist to evaluate bone age (BA). It evaluates 13 bones, including the ulna, radius, and 11 short bones in fingers 1, 3, and 5. Bone morphology, density scores, and textural features serve as critical parameters for this algorithm to discern and distinguish bone structures. The radiograph analysis is segmented into three successive layers. Initially, the software identifies bones of interest by applying active appearance models. Subsequently, it determines and verifies the bone age for each identified bone. In the final stage, the software converts the computed BA values into GP or TW BA values [10]. Recently, another bone age application called Physis® (developed by 16-bit AI™, Toronto, Canada) has been introduced [11]. It analyses left-hand and wrist radiographs, providing the predicted bone age. This application, which was the winner of the 2017 RSNA Paediatric Bone Age Challenge, attained a concordance correlation coefficient (CCC) of 0.991 when compared to the ground truth determined by radiologists along with a mean absolute difference of 4.265 months [12].

In our study, we aimed to compare agreement between measurements obtained using the standard GP method, 16-bit AItm software (free version begin 2022), and BoneXpert® system.