Application of convolutional neural networks for distal radio-ulnar fracture detection on plain radiographs in the emergency room
Objective Recent studies have suggested that deep-learning models can satisfactorily assist in fracture diagnosis. We aimed to evaluate the performance of two of such models in wrist fracture detection. Methods We collected image data of patients who visited with wrist trauma at the emergency depart...
Main Authors: | , , , , , , , , |
---|---|
Format: | Article |
Language: | English |
Published: |
The Korean Society of Emergency Medicine
2021-06-01
|
Series: | Clinical and Experimental Emergency Medicine |
Subjects: | |
Online Access: | http://ceemjournal.org/upload/pdf/ceem-20-091.pdf |
_version_ | 1797897180463235072 |
---|---|
author | Min Woong Kim Jaewon Jung Se Jin Park Young Sun Park Jeong Hyeon Yi Won Seok Yang Jin Hyuck Kim Bum-Joo Cho Sang Ook Ha |
author_facet | Min Woong Kim Jaewon Jung Se Jin Park Young Sun Park Jeong Hyeon Yi Won Seok Yang Jin Hyuck Kim Bum-Joo Cho Sang Ook Ha |
author_sort | Min Woong Kim |
collection | DOAJ |
description | Objective Recent studies have suggested that deep-learning models can satisfactorily assist in fracture diagnosis. We aimed to evaluate the performance of two of such models in wrist fracture detection. Methods We collected image data of patients who visited with wrist trauma at the emergency department. A dataset extracted from January 2018 to May 2020 was split into training (90%) and test (10%) datasets, and two types of convolutional neural networks (i.e., DenseNet-161 and ResNet-152) were trained to detect wrist fractures. Gradient-weighted class activation mapping was used to highlight the regions of radiograph scans that contributed to the decision of the model. Performance of the convolutional neural network models was evaluated using the area under the receiver operating characteristic curve. Results For model training, we used 4,551 radiographs from 798 patients and 4,443 radiographs from 1,481 patients with and without fractures, respectively. The remaining 10% (300 radiographs from 100 patients with fractures and 690 radiographs from 230 patients without fractures) was used as a test dataset. The sensitivity, specificity, positive predictive value, negative predictive value, and accuracy of DenseNet-161 and ResNet-152 in the test dataset were 90.3%, 90.3%, 80.3%, 95.6%, and 90.3% and 88.6%, 88.4%, 76.9%, 94.7%, and 88.5%, respectively. The area under the receiver operating characteristic curves of DenseNet-161 and ResNet-152 for wrist fracture detection were 0.962 and 0.947, respectively. Conclusion We demonstrated that DenseNet-161 and ResNet-152 models could help detect wrist fractures in the emergency room with satisfactory performance. |
first_indexed | 2024-04-10T07:53:28Z |
format | Article |
id | doaj.art-c242b47aa649493eb0b2e41b6819caa1 |
institution | Directory Open Access Journal |
issn | 2383-4625 |
language | English |
last_indexed | 2024-04-10T07:53:28Z |
publishDate | 2021-06-01 |
publisher | The Korean Society of Emergency Medicine |
record_format | Article |
series | Clinical and Experimental Emergency Medicine |
spelling | doaj.art-c242b47aa649493eb0b2e41b6819caa12023-02-23T07:04:14ZengThe Korean Society of Emergency MedicineClinical and Experimental Emergency Medicine2383-46252021-06-018212012710.15441/ceem.20.091337Application of convolutional neural networks for distal radio-ulnar fracture detection on plain radiographs in the emergency roomMin Woong Kim0Jaewon Jung1Se Jin Park2Young Sun Park3Jeong Hyeon Yi4Won Seok Yang5Jin Hyuck Kim6Bum-Joo Cho7Sang Ook Ha8 Department of Emergency Medicine, Hallym University Sacred Heart Hospital, Hallym University Medical Center, Anyang, Korea Medical Artificial Intelligence Center, Hallym University Sacred Heart Hospital, Hallym University Medical Center, Anyang, Korea Medical Artificial Intelligence Center, Hallym University Sacred Heart Hospital, Hallym University Medical Center, Anyang, Korea Department of Emergency Medicine, Hallym University Sacred Heart Hospital, Hallym University Medical Center, Anyang, Korea Department of Emergency Medicine, Hallym University Sacred Heart Hospital, Hallym University Medical Center, Anyang, Korea Department of Emergency Medicine, Hallym University Sacred Heart Hospital, Hallym University Medical Center, Anyang, Korea Department of Neurology, Hallym University Sacred Heart Hospital, Hallym University Medical Center, Anyang, Korea Medical Artificial Intelligence Center, Hallym University Sacred Heart Hospital, Hallym University Medical Center, Anyang, Korea Department of Emergency Medicine, Hallym University Sacred Heart Hospital, Hallym University Medical Center, Anyang, KoreaObjective Recent studies have suggested that deep-learning models can satisfactorily assist in fracture diagnosis. We aimed to evaluate the performance of two of such models in wrist fracture detection. Methods We collected image data of patients who visited with wrist trauma at the emergency department. A dataset extracted from January 2018 to May 2020 was split into training (90%) and test (10%) datasets, and two types of convolutional neural networks (i.e., DenseNet-161 and ResNet-152) were trained to detect wrist fractures. Gradient-weighted class activation mapping was used to highlight the regions of radiograph scans that contributed to the decision of the model. Performance of the convolutional neural network models was evaluated using the area under the receiver operating characteristic curve. Results For model training, we used 4,551 radiographs from 798 patients and 4,443 radiographs from 1,481 patients with and without fractures, respectively. The remaining 10% (300 radiographs from 100 patients with fractures and 690 radiographs from 230 patients without fractures) was used as a test dataset. The sensitivity, specificity, positive predictive value, negative predictive value, and accuracy of DenseNet-161 and ResNet-152 in the test dataset were 90.3%, 90.3%, 80.3%, 95.6%, and 90.3% and 88.6%, 88.4%, 76.9%, 94.7%, and 88.5%, respectively. The area under the receiver operating characteristic curves of DenseNet-161 and ResNet-152 for wrist fracture detection were 0.962 and 0.947, respectively. Conclusion We demonstrated that DenseNet-161 and ResNet-152 models could help detect wrist fractures in the emergency room with satisfactory performance.http://ceemjournal.org/upload/pdf/ceem-20-091.pdfwristfractures, bonedeep learningneural networks, computer |
spellingShingle | Min Woong Kim Jaewon Jung Se Jin Park Young Sun Park Jeong Hyeon Yi Won Seok Yang Jin Hyuck Kim Bum-Joo Cho Sang Ook Ha Application of convolutional neural networks for distal radio-ulnar fracture detection on plain radiographs in the emergency room Clinical and Experimental Emergency Medicine wrist fractures, bone deep learning neural networks, computer |
title | Application of convolutional neural networks for distal radio-ulnar fracture detection on plain radiographs in the emergency room |
title_full | Application of convolutional neural networks for distal radio-ulnar fracture detection on plain radiographs in the emergency room |
title_fullStr | Application of convolutional neural networks for distal radio-ulnar fracture detection on plain radiographs in the emergency room |
title_full_unstemmed | Application of convolutional neural networks for distal radio-ulnar fracture detection on plain radiographs in the emergency room |
title_short | Application of convolutional neural networks for distal radio-ulnar fracture detection on plain radiographs in the emergency room |
title_sort | application of convolutional neural networks for distal radio ulnar fracture detection on plain radiographs in the emergency room |
topic | wrist fractures, bone deep learning neural networks, computer |
url | http://ceemjournal.org/upload/pdf/ceem-20-091.pdf |
work_keys_str_mv | AT minwoongkim applicationofconvolutionalneuralnetworksfordistalradioulnarfracturedetectiononplainradiographsintheemergencyroom AT jaewonjung applicationofconvolutionalneuralnetworksfordistalradioulnarfracturedetectiononplainradiographsintheemergencyroom AT sejinpark applicationofconvolutionalneuralnetworksfordistalradioulnarfracturedetectiononplainradiographsintheemergencyroom AT youngsunpark applicationofconvolutionalneuralnetworksfordistalradioulnarfracturedetectiononplainradiographsintheemergencyroom AT jeonghyeonyi applicationofconvolutionalneuralnetworksfordistalradioulnarfracturedetectiononplainradiographsintheemergencyroom AT wonseokyang applicationofconvolutionalneuralnetworksfordistalradioulnarfracturedetectiononplainradiographsintheemergencyroom AT jinhyuckkim applicationofconvolutionalneuralnetworksfordistalradioulnarfracturedetectiononplainradiographsintheemergencyroom AT bumjoocho applicationofconvolutionalneuralnetworksfordistalradioulnarfracturedetectiononplainradiographsintheemergencyroom AT sangookha applicationofconvolutionalneuralnetworksfordistalradioulnarfracturedetectiononplainradiographsintheemergencyroom |