Doctor Imitator: A Graph-Based Bone Age Assessment Framework Using Hand Radiographs

Abstract

Bone age assessment is challenging in clinical practice due to the complicated bone age assessment process. Current automatic bone age assessment methods were designed with rare consideration of the diagnostic logistics and thus may yield certain uninterpretable hidden states and outputs. Consequently, doctors can find it hard to cooperate with such models harmoniously because it is difficult to check the correctness of the model predictions. In this work, we propose a new graph-based deep learning framework for bone age assessment with hand radiographs, called Doctor Imitator (DI). The architecture of DI is designed to learn the diagnostic logistics of doctors using the scoring methods (e.g., the Tanner-Whitehouse method) for bone age assessment. Specifically, the convolutions of DI capture the local features of the anatomical regions of interest (ROIs) on hand radiographs and predict the ROI scores by our proposed Anatomy-based Group Convolution, summing up for bone age prediction. Besides, we develop a novel Dual Graph-based Attention module to compute patient-specific attention for ROI features and context attention for ROI scores. As far as we know, DI is the first automatic bone age assessment framework following the scoring methods without fully supervised hand radiographs. Experiments on hand radiographs with only bone age supervision verify that DI can achieve excellent performance with sparse parameters and provide more interpretability.

Bone age assessment based on deep learning architecture

“…Deep learning has grown in popularity in recent years due to the ability to learn features automatically. convolutional neural networks (CNNs) typically have three-layer types: convolutional layer, pooling layer, and fully connected layer see in Figure 3 [24]. The convolutional layer is responsible for computing the weighted sum, including a bias value into the weighted sum, and then using a function of activation known as the rectifier linear unit (ReLu) to the addition result., which is described using (1).…”

“…Figure 3. Convolutional network architecture (CNN) in general [24] In this work, we evaluate the performance of several designs and compare the results. This includes a design model comprised of 10 learnable weighted layers: 7 convolutional blocks and 3 fully-connected layers.…”

A Neural Network for High-Precise and Well-Interpretable Electrocardiogram Classification

“…Neural networks have demonstrated significant superiority over traditional machine learning methods in various domains, including image recognition [10], speech recognition [11], natural language processing [12], [13], medical image analysis [14]- [16], and point cloud data analysis [17], [18]. Given the global repository of over 300 million recorded ECGs [19], these vast datasets provide an ideal foundation for training large-scale neural networks.…”