Classification of hand‐wrist maturity level based on similarity matching

Abstract

Judging the maturity level of each hand-wrist reference bone is the core issue in bone age assessment. Relying on the superiority of convolutional neural networks in feature representation, deep learning is widely studied for the automatic bone age assessment. However, an efficient but complex deep learning network requests a large dataset with bone-maturitylevel labels for training, restricting its large-scale application in bone maturity classification. For this reason, we transform the bone-maturity-level classification problem into the similarity matching problem. Also, we propose a general structure based on Siamese network by merging two inputs into a two-channel input and introducing a dual attention mechanism, to create an Attentional Two-Channel Network (ATC-Net). This paper takes the intermediate phalanges III as an example to assess the performance of the similarity matching method and the ATC-Net. Experiments show that our method can perform better on small datasets, which effectively makes up for the data shortage problem. The ATC-Net used for classification significantly reduces the evaluation time compared with other classical networks. It reduces the time of assessing one sample by about 49% as compared to VGG-16. And more importantly, it achieves the highest classification accuracy of 92.74% among all investigated networks.

Using artificial intelligence models to evaluate envisaged points initially: A pilot study

“…The developed classification model achieved a classification accuracy of 92.74%, significantly reduced the time of assessing one sample by about 49% as compared to VGG-16 and can be used to assess the bone age based on CHN, TW3 or other scoring methods. 15…”

“…However, the CNNs require a large data set with good quality labeling and insufficient data seriously affects the feature extraction and classification ability of the model. 15…”

“…DL algorithms can automatically learn from a large data set and drive many AI applications without or limited human intervention. 15 Convolutional neural networks (CNNs) are specific DL algorithms which are inspired by the visual cortex in the brain and primarily used in computer vision and image classification tasks. Basic architecture of the CNNs is based on classical NN architecture with additional layers such as convolutional and pooling layers.…”

“…The convolutional layers in the CNNs architecture serve as feature extractors. 15,16 At the end of the 1980s, CNNs was the focus of automatic bone age assessment researches, and in 1990s, computer-assisted bone age scoring systems were available to the operators. However, the CNNs require a large data set with good quality labeling and insufficient data seriously affects the feature extraction and classification ability of the model.…”

Hamate classification method based on feature‐enhanced residual network and probabilistic joint judgment

“…In order to provide a more scientific assessment and timely medical intervention of adolescent growth and development, bone age is commonly used in clinical practice as the main criterion [2]. Bone age reflects the maturity of growth and development through the characteristics of skeletal development shown in X‐ray images [3, 4]. As an independent indicator of growth, bone age is not dependent on age or growth rate [5].…”