Multi-scale multi-reception attention network for bone age assessment in X-ray images

Area of Science:

• Radiology
• Medical Imaging
• Artificial Intelligence

Background:

• Bone age assessment is crucial for evaluating skeletal maturity.
• Hand radiograph analysis for bone age often includes redundant information, complicating analysis.
• Current deep learning methods for bone age assessment can be complex and require extensive annotations.

Purpose of the Study:

• To develop an efficient and accurate deep learning model for bone age assessment.
• To improve feature representation in hand radiographs by focusing on key regions.
• To reduce the reliance on complex network structures and additional annotations.

Main Methods:

• A Multi-scale Multi-reception Attention Net (MMANet) was developed, integrating a Multi-scale Multi-reception Complement Attention (MMCA) module.
• A ResNet backbone was utilized to enhance feature representation and suppress background noise.
• The model employs a graph attention module to focus on critical anatomical regions.

Main Results:

• MMANet achieved a mean absolute error (MAE) of 3.88 months on the RSNA 2017 Paediatric Bone Age Challenge dataset.
• The proposed method demonstrated superior performance compared to the state-of-the-art, with an MAE of 3.91 months.
• MMANet accurately detected key regions without explicit anatomical information modeling.

Conclusions:

• MMANet offers a highly accurate and efficient approach to bone age assessment from radiographs.
• The model's performance surpasses current state-of-the-art methods, particularly in its reduced need for annotations.
• This work provides a valuable tool for pediatric bone maturity estimation in clinical practice.