تخمین سن استخوان از طریق آنالیز اشعه ایکس دست با مدل ترانسفورماتور بینایی
کد: G-1490
نویسندگان: Armin Abdollahi * ℗, Maliheh Sabeti, Reza Boostani
زمان بندی: زمان بندی نشده!
برچسب: پردازش تصاویر پزشکی
دانلود: دانلود پوستر
خلاصه مقاله:
خلاصه مقاله
Background and Aims: Bone age estimation is a critical task in pediatric medicine, aiding in the diagnosis of growth disorders and treatment planning. Traditional methods, such as the Greulich-Pyle and Tanner-Whitehouse techniques, depend on radiologists’ subjective interpretation of hand X-rays, leading to variability and time inefficiency. With recent advances in deep learning, especially Transformer-based models, automated bone age estimation has gained considerable attention. This study aims to propose an accurate and efficient model for bone age prediction using the Vision Transformer (ViT) architecture. Method: The study utilized the Atlas dataset, consisting of 1,390 hand X-ray images of children aged from infancy to 18 years. All images were resized to 512×512 pixels, and data augmentation techniques were applied to enhance generalizability and handle class imbalance. The ViT model was trained using the Adam optimizer, and Mean Squared Error (MSE) was used as the loss function. Evaluation metrics included Mean Absolute Error (MAE) and overall classification accuracy. Results: The proposed Vision Transformer model achieved a classification accuracy of 92% and a Mean Absolute Error of 3.2 months. Compared to conventional Convolutional Neural Network (CNN) models applied to the same dataset, the ViT architecture demonstrated superior performance by effectively capturing intricate visual patterns in X-ray images. Conclusion: The ViT-based model shows strong potential in improving the accuracy and reliability of bone age estimation. By minimizing human error and reducing dependency on manual assessment, this approach offers a valuable tool for pediatric diagnostics, enhancing both clinical outcomes and workflow efficiency.
کلمات کلیدی
Bone Age Estimation, Transformers, DeepLearning, Healthcare
