Severity Classification of Conjunctival Hyperaemia by Deep Neural Network Ensembles
Conjunctival hyperaemia is a common clinical ophthalmological finding and can be a symptom of various ocular disorders. Although several severity classification criteria have been proposed, none include objective severity criteria. Neural networks and deep learning have been utilised in ophthalmolog...
Saved in:
| Main Authors: | , , , , , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
Wiley
2019-01-01
|
| Series: | Journal of Ophthalmology |
| Online Access: | http://dx.doi.org/10.1155/2019/7820971 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| _version_ | 1832561688752160768 |
|---|---|
| author | Hiroki Masumoto Hitoshi Tabuchi Tsuyoshi Yoneda Shunsuke Nakakura Hideharu Ohsugi Tamaki Sumi Atsuki Fukushima |
| author_facet | Hiroki Masumoto Hitoshi Tabuchi Tsuyoshi Yoneda Shunsuke Nakakura Hideharu Ohsugi Tamaki Sumi Atsuki Fukushima |
| author_sort | Hiroki Masumoto |
| collection | DOAJ |
| description | Conjunctival hyperaemia is a common clinical ophthalmological finding and can be a symptom of various ocular disorders. Although several severity classification criteria have been proposed, none include objective severity criteria. Neural networks and deep learning have been utilised in ophthalmology, but not for the purpose of classifying the severity of conjunctival hyperaemia objectively. To develop a conjunctival hyperaemia grading software, we used 3700 images as the training data and 923 images as the validation test data. We trained the nine neural network models and validated the performance of these networks. We finally chose the best combination of these networks. The DenseNet201 model was the best individual model. The combination of the DenseNet201, DenseNet121, VGG19, and ResNet50 were the best model. The correlation between the multimodel responses, and the vessel-area occupied was 0.737 (p<0.01). This system could be as accurate and comprehensive as specialists but would be significantly faster and consistent with objective values. |
| format | Article |
| id | doaj-art-1998dc5a67194af59dadd791787cde0d |
| institution | Kabale University |
| issn | 2090-004X 2090-0058 |
| language | English |
| publishDate | 2019-01-01 |
| publisher | Wiley |
| record_format | Article |
| series | Journal of Ophthalmology |
| spelling | doaj-art-1998dc5a67194af59dadd791787cde0d2025-02-03T01:24:24ZengWileyJournal of Ophthalmology2090-004X2090-00582019-01-01201910.1155/2019/78209717820971Severity Classification of Conjunctival Hyperaemia by Deep Neural Network EnsemblesHiroki Masumoto0Hitoshi Tabuchi1Tsuyoshi Yoneda2Shunsuke Nakakura3Hideharu Ohsugi4Tamaki Sumi5Atsuki Fukushima6Department of Ophthalmology, Tsukazaki Hospital, Himeji, JapanDepartment of Ophthalmology, Tsukazaki Hospital, Himeji, JapanDepartment of Sensory Science, Kawasaki University of Medical Welfare, Kurashiki, JapanDepartment of Ophthalmology, Tsukazaki Hospital, Himeji, JapanDepartment of Ophthalmology, Tsukazaki Hospital, Himeji, JapanDepartment of Ophthalmology and Visual Science, Kochi Medical School, Nankoku, JapanDepartment of Ophthalmology and Visual Science, Kochi Medical School, Nankoku, JapanConjunctival hyperaemia is a common clinical ophthalmological finding and can be a symptom of various ocular disorders. Although several severity classification criteria have been proposed, none include objective severity criteria. Neural networks and deep learning have been utilised in ophthalmology, but not for the purpose of classifying the severity of conjunctival hyperaemia objectively. To develop a conjunctival hyperaemia grading software, we used 3700 images as the training data and 923 images as the validation test data. We trained the nine neural network models and validated the performance of these networks. We finally chose the best combination of these networks. The DenseNet201 model was the best individual model. The combination of the DenseNet201, DenseNet121, VGG19, and ResNet50 were the best model. The correlation between the multimodel responses, and the vessel-area occupied was 0.737 (p<0.01). This system could be as accurate and comprehensive as specialists but would be significantly faster and consistent with objective values.http://dx.doi.org/10.1155/2019/7820971 |
| spellingShingle | Hiroki Masumoto Hitoshi Tabuchi Tsuyoshi Yoneda Shunsuke Nakakura Hideharu Ohsugi Tamaki Sumi Atsuki Fukushima Severity Classification of Conjunctival Hyperaemia by Deep Neural Network Ensembles Journal of Ophthalmology |
| title | Severity Classification of Conjunctival Hyperaemia by Deep Neural Network Ensembles |
| title_full | Severity Classification of Conjunctival Hyperaemia by Deep Neural Network Ensembles |
| title_fullStr | Severity Classification of Conjunctival Hyperaemia by Deep Neural Network Ensembles |
| title_full_unstemmed | Severity Classification of Conjunctival Hyperaemia by Deep Neural Network Ensembles |
| title_short | Severity Classification of Conjunctival Hyperaemia by Deep Neural Network Ensembles |
| title_sort | severity classification of conjunctival hyperaemia by deep neural network ensembles |
| url | http://dx.doi.org/10.1155/2019/7820971 |
| work_keys_str_mv | AT hirokimasumoto severityclassificationofconjunctivalhyperaemiabydeepneuralnetworkensembles AT hitoshitabuchi severityclassificationofconjunctivalhyperaemiabydeepneuralnetworkensembles AT tsuyoshiyoneda severityclassificationofconjunctivalhyperaemiabydeepneuralnetworkensembles AT shunsukenakakura severityclassificationofconjunctivalhyperaemiabydeepneuralnetworkensembles AT hideharuohsugi severityclassificationofconjunctivalhyperaemiabydeepneuralnetworkensembles AT tamakisumi severityclassificationofconjunctivalhyperaemiabydeepneuralnetworkensembles AT atsukifukushima severityclassificationofconjunctivalhyperaemiabydeepneuralnetworkensembles |