Development of skeletal bone age assessment model using multi-convolutional with multi-attention based residual densenet with optimal tuning

Assessment of skeletal bone age is an essential method and extensively applied across medical and scientific disciplines to estimate the developmental progression of an individual’s bones. Unlike chronological age, the skeleton bone age provides a more accurate outcome that measures the individual’s biological and physiological maturity. Based on this, a bone age assessment is essential for assessing the overall growth, especially in specialized applications, such as orthodontics. However, the process of determining bone age is complex and time-consuming. It usually requires the examination of hand bone X-rays and comparison with established reference standards, a task that demands expert analysis. This manual procedure is susceptible to inconsistencies, both between different evaluators (inter-observer differences) and within the same evaluator over time (intra-observer differences), resulting in variable outcomes. Therefore, a fully automated, reliable, and accurate model is developed for assessing skeletal bone age. At first, the relevant input images are collected from standard source. These images are given to the developed Multi-Convolutional with Multi-Attention based Residual Densenet with Optimal Tuning Model for performing the skeletal bone age assessment. Here, each convolution carries each attention type, including spatial attention, region attention and sparse attention mechanism. Additionally, the parameters of the MC-MARDNet are optimized by utilizing the Modified Random Attribute-based Flower Fertilization Optimization (MRA-FFO). This model supports in monitoring the children’s physical development and also identify the growth hormone deficiency. The performance estimation of the developed model shows resultant outcomes when compared with traditional methodologies.