Deep learning approach based on a patch residual for pediatric supracondylar subtle fracture detection

Supracondylar humerus fractures in children are among the most common elbow fractures in pediatrics. However, their diagnosis can be particularly challenging due to the anatomical characteristics and imaging features of the pediatric skeleton. In recent years, convolutional neural networks (CNNs) ha...

Full description

Saved in:
Bibliographic Details
Main Authors: Qingming Ye, Zhilu Wang, Yi Lou, Yang Yang, Jue Hou, Zheng Liu, Weiguang Liu, Jiayu Li
Format: Article
Language:English
Published: Association of Basic Medical Sciences of Federation of Bosnia and Herzegovina 2025-01-01
Series:Biomolecules & Biomedicine
Subjects:
Online Access:https://www.bjbms.org/ojs/index.php/bjbms/article/view/11341
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1850095804128165888
author Qingming Ye
Zhilu Wang
Yi Lou
Yang Yang
Jue Hou
Zheng Liu
Weiguang Liu
Jiayu Li
author_facet Qingming Ye
Zhilu Wang
Yi Lou
Yang Yang
Jue Hou
Zheng Liu
Weiguang Liu
Jiayu Li
author_sort Qingming Ye
collection DOAJ
description Supracondylar humerus fractures in children are among the most common elbow fractures in pediatrics. However, their diagnosis can be particularly challenging due to the anatomical characteristics and imaging features of the pediatric skeleton. In recent years, convolutional neural networks (CNNs) have achieved notable success in medical image analysis, though their performance typically relies on large-scale, high-quality labeled datasets. Unfortunately, labeled samples for pediatric supracondylar fractures are scarce and difficult to obtain. To address this issue, this paper introduces a deep learning-based multiscale patch residual network (MPR) for the automatic detection and localization of subtle pediatric supracondylar fractures. The MPR framework combines a CNN for automatic feature extraction with a multiscale generative adversarial network (GAN) to model skeletal integrity using healthy samples. By leveraging healthy images to learn the normal skeletal distribution, the approach reduces the dependency on labeled fracture data and effectively addresses the challenges posed by limited pediatric datasets. Datasets from two different hospitals were used, with data augmentation techniques applied during both training and validation. On an independent test set, the proposed model achieves an accuracy of 90.5%, with 89% sensitivity, 92% specificity, and an F1 score of 0.906—outperforming the diagnostic accuracy of emergency medicine physicians and approaching that of pediatric radiologists. Furthermore, the model demonstrates a fast inference speed of 1.1 seconds per sheet, underscoring its substantial potential for clinical application.
format Article
id doaj-art-18f5d40935db4044b01926c882cc2bfa
institution DOAJ
issn 2831-0896
2831-090X
language English
publishDate 2025-01-01
publisher Association of Basic Medical Sciences of Federation of Bosnia and Herzegovina
record_format Article
series Biomolecules & Biomedicine
spelling doaj-art-18f5d40935db4044b01926c882cc2bfa2025-08-20T02:41:21ZengAssociation of Basic Medical Sciences of Federation of Bosnia and HerzegovinaBiomolecules & Biomedicine2831-08962831-090X2025-01-0110.17305/bb.2024.11341Deep learning approach based on a patch residual for pediatric supracondylar subtle fracture detectionQingming Ye0https://orcid.org/0009-0004-7745-9757Zhilu Wang1Yi Lou2Yang Yang3Jue Hou4Zheng Liu5Weiguang Liu6Jiayu Li7Zhejiang Sci-Tech University, Hangzhou, ChinaZhejiang Sci-Tech University, Hangzhou, ChinaHangzhou Children's Hospital, Hangzhou, China Zhejiang Sci-Tech University, Hangzhou, ChinaZhejiang Sci-Tech University, Hangzhou, ChinaZhejiang Sci-Tech University, Hangzhou, ChinaHangzhou Children's Hospital, Hangzhou, China First Teaching Hospital of Tianjin University of Traditional Chinese Medicine, Tianjin, ChinaSupracondylar humerus fractures in children are among the most common elbow fractures in pediatrics. However, their diagnosis can be particularly challenging due to the anatomical characteristics and imaging features of the pediatric skeleton. In recent years, convolutional neural networks (CNNs) have achieved notable success in medical image analysis, though their performance typically relies on large-scale, high-quality labeled datasets. Unfortunately, labeled samples for pediatric supracondylar fractures are scarce and difficult to obtain. To address this issue, this paper introduces a deep learning-based multiscale patch residual network (MPR) for the automatic detection and localization of subtle pediatric supracondylar fractures. The MPR framework combines a CNN for automatic feature extraction with a multiscale generative adversarial network (GAN) to model skeletal integrity using healthy samples. By leveraging healthy images to learn the normal skeletal distribution, the approach reduces the dependency on labeled fracture data and effectively addresses the challenges posed by limited pediatric datasets. Datasets from two different hospitals were used, with data augmentation techniques applied during both training and validation. On an independent test set, the proposed model achieves an accuracy of 90.5%, with 89% sensitivity, 92% specificity, and an F1 score of 0.906—outperforming the diagnostic accuracy of emergency medicine physicians and approaching that of pediatric radiologists. Furthermore, the model demonstrates a fast inference speed of 1.1 seconds per sheet, underscoring its substantial potential for clinical application. https://www.bjbms.org/ojs/index.php/bjbms/article/view/11341Deep learningpediatric supracondylar subtle fracturefracture detectionsmall data
spellingShingle Qingming Ye
Zhilu Wang
Yi Lou
Yang Yang
Jue Hou
Zheng Liu
Weiguang Liu
Jiayu Li
Deep learning approach based on a patch residual for pediatric supracondylar subtle fracture detection
Biomolecules & Biomedicine
Deep learning
pediatric supracondylar subtle fracture
fracture detection
small data
title Deep learning approach based on a patch residual for pediatric supracondylar subtle fracture detection
title_full Deep learning approach based on a patch residual for pediatric supracondylar subtle fracture detection
title_fullStr Deep learning approach based on a patch residual for pediatric supracondylar subtle fracture detection
title_full_unstemmed Deep learning approach based on a patch residual for pediatric supracondylar subtle fracture detection
title_short Deep learning approach based on a patch residual for pediatric supracondylar subtle fracture detection
title_sort deep learning approach based on a patch residual for pediatric supracondylar subtle fracture detection
topic Deep learning
pediatric supracondylar subtle fracture
fracture detection
small data
url https://www.bjbms.org/ojs/index.php/bjbms/article/view/11341
work_keys_str_mv AT qingmingye deeplearningapproachbasedonapatchresidualforpediatricsupracondylarsubtlefracturedetection
AT zhiluwang deeplearningapproachbasedonapatchresidualforpediatricsupracondylarsubtlefracturedetection
AT yilou deeplearningapproachbasedonapatchresidualforpediatricsupracondylarsubtlefracturedetection
AT yangyang deeplearningapproachbasedonapatchresidualforpediatricsupracondylarsubtlefracturedetection
AT juehou deeplearningapproachbasedonapatchresidualforpediatricsupracondylarsubtlefracturedetection
AT zhengliu deeplearningapproachbasedonapatchresidualforpediatricsupracondylarsubtlefracturedetection
AT weiguangliu deeplearningapproachbasedonapatchresidualforpediatricsupracondylarsubtlefracturedetection
AT jiayuli deeplearningapproachbasedonapatchresidualforpediatricsupracondylarsubtlefracturedetection