Bone Age Estimation of Pediatrics by Analyzing Hand X-Rays Using Deep Learning Technique

The determination of bone age is critical for detecting metabolic and endocrine problems in a child's development. It provides important insights on the rate of structural and biological development, which frequently differs compared to the chronological age determined at birth. This study presents a completely automated deep learning technique for correctly determining bone age from X-ray images of the hand. The dataset used for training and evaluation is derived from the Radiological Society of North America's 2017 Pediatric Bone Age Challenge, which comprises left hand X-ray pictures annotated with gender and age information. To determine bone age precisely, we use a transfer learning technique using the pre-trained Xception model. By fine-tuning the neural network on the bone age dataset, it was possible to capture complicated patterns and attributes peculiar to bone development. With a mean absolute error (MAE) of 1.52 months, the experimental results show a remarkable level of convergence between the predicted age and the actual chronological age. The therapeutic significance of our suggested technique arises from its potential to be a beneficial tool to assist medical practitioners in more correctly and effectively determining bone age. The incorporation of AI-based autonomous bone age assessment can help reduce the diagnostic process and assist in early identification of developmental anomalies, ultimately contributing to improved pediatric healthcare.

I. Introduction

For detecting and treating endocrine and growth abnormalities, the process of determining skeletal maturity known as “bone age assessment” (BAA) is used [1]. BAA is an alternative method for determining age when accurate birth records aren't kept. The difference between bone age and chronological age implies skeletal development issues. [2]. In the standard BAA procedure, radiologists look at x-ray of the left hand that shows fingers and wrist. In practice, the traditional Greulich and Pyle (GP) technique [3] and Tanner Whitehouse method (TW2) [4] are often utilized. The first method measures bone age by comparing the patient's hand X-ray image to an atlas. [3]. The second method uses significant areas of interest, like hand bones, to determine the age of the bones [4].