Matches in SemOpenAlex for { <https://semopenalex.org/work/W4386868435> ?p ?o ?g. }
Showing items 1 to 94 of
94
with 100 items per page.
- W4386868435 abstract "Chest X-ray (CXR) is essential for physicians to diagnose lung diseases in clinical medicine. With the development of computer and deep learning techniques, creating CXR datasets to train convolutional neural networks (CNNs) has become a popular research topic. However, there are still challenges in creating datasets due to class imbalance and the ratio setting of datasets. This study investigated the impact of class imbalance and the ratio of training, validation, and test sets on CNNs classification performance by optimizing the dataset configuration. This was achieved by directly modifying the ratio, applying oversampling based on the adaptive contrast enhancement (ACE) algorithm, and random undersampling to balance the dataset classes, followed by modifying the ratio again. Therefore, seven datasets were obtained, which were utilized to individually train four CNNs based on transfer learning and fine-tuning techniques. Evaluation metrics based on the confusion matrix were utilized to demonstrate the enhanced classification performance of CNNs. The results indicated that at least 17% of the accuracy of CNNs trained by the dataset with modified ratios was improved as compared with the dataset with an unreasonable initial configuration. Additionally, the overall evaluation metrics of CNNs were further improved by balancing the dataset classes and modifying the ratios. ChexNet demonstrated the best classification performance among the four CNNs, as evidenced by the area under the receiver operating characteristic (ROC) curves (AUC). Furthermore, ChexNet was trained by a class balanced dataset with a ratio of 7:2:1, resulting in the best evaluation metrics, including F1 Score, balanced accuracy, accuracy, recall, specificity, and precision, which were 97.81%, 97.78%, 97.78%, 98.73%, 96.84%, and 96.89%, respectively. Therefore, optimizing dataset configuration can effectively improve the performance of CNNs, providing empirical guidance for researchers in creating datasets." @default.
- W4386868435 created "2023-09-20" @default.
- W4386868435 creator A5045519742 @default.
- W4386868435 creator A5064042477 @default.
- W4386868435 creator A5075482179 @default.
- W4386868435 date "2023-04-21" @default.
- W4386868435 modified "2023-09-27" @default.
- W4386868435 title "Optimizing Dataset Configuration for Improving Convolutional Neural Networks Performance in Chest X-ray Image Classification" @default.
- W4386868435 cites W1677182931 @default.
- W4386868435 cites W1941659294 @default.
- W4386868435 cites W2057298051 @default.
- W4386868435 cites W2097117768 @default.
- W4386868435 cites W2101807845 @default.
- W4386868435 cites W2152772232 @default.
- W4386868435 cites W2165844808 @default.
- W4386868435 cites W2194775991 @default.
- W4386868435 cites W2301358467 @default.
- W4386868435 cites W2562319768 @default.
- W4386868435 cites W2585770658 @default.
- W4386868435 cites W2592929672 @default.
- W4386868435 cites W2618530766 @default.
- W4386868435 cites W2734760437 @default.
- W4386868435 cites W2767106145 @default.
- W4386868435 cites W2788633781 @default.
- W4386868435 cites W2891756914 @default.
- W4386868435 cites W2924911266 @default.
- W4386868435 cites W2963446712 @default.
- W4386868435 cites W3013277995 @default.
- W4386868435 cites W3025815763 @default.
- W4386868435 cites W3105081694 @default.
- W4386868435 cites W3106539405 @default.
- W4386868435 cites W4283793640 @default.
- W4386868435 doi "https://doi.org/10.1109/icsp58490.2023.10248725" @default.
- W4386868435 hasPublicationYear "2023" @default.
- W4386868435 type Work @default.
- W4386868435 citedByCount "0" @default.
- W4386868435 crossrefType "proceedings-article" @default.
- W4386868435 hasAuthorship W4386868435A5045519742 @default.
- W4386868435 hasAuthorship W4386868435A5064042477 @default.
- W4386868435 hasAuthorship W4386868435A5075482179 @default.
- W4386868435 hasConcept C115961682 @default.
- W4386868435 hasConcept C119857082 @default.
- W4386868435 hasConcept C124101348 @default.
- W4386868435 hasConcept C134306372 @default.
- W4386868435 hasConcept C136536468 @default.
- W4386868435 hasConcept C138602881 @default.
- W4386868435 hasConcept C148524875 @default.
- W4386868435 hasConcept C153180895 @default.
- W4386868435 hasConcept C154945302 @default.
- W4386868435 hasConcept C177148314 @default.
- W4386868435 hasConcept C197323446 @default.
- W4386868435 hasConcept C2776257435 @default.
- W4386868435 hasConcept C2777212361 @default.
- W4386868435 hasConcept C31258907 @default.
- W4386868435 hasConcept C33923547 @default.
- W4386868435 hasConcept C41008148 @default.
- W4386868435 hasConcept C58471807 @default.
- W4386868435 hasConcept C75294576 @default.
- W4386868435 hasConcept C81363708 @default.
- W4386868435 hasConceptScore W4386868435C115961682 @default.
- W4386868435 hasConceptScore W4386868435C119857082 @default.
- W4386868435 hasConceptScore W4386868435C124101348 @default.
- W4386868435 hasConceptScore W4386868435C134306372 @default.
- W4386868435 hasConceptScore W4386868435C136536468 @default.
- W4386868435 hasConceptScore W4386868435C138602881 @default.
- W4386868435 hasConceptScore W4386868435C148524875 @default.
- W4386868435 hasConceptScore W4386868435C153180895 @default.
- W4386868435 hasConceptScore W4386868435C154945302 @default.
- W4386868435 hasConceptScore W4386868435C177148314 @default.
- W4386868435 hasConceptScore W4386868435C197323446 @default.
- W4386868435 hasConceptScore W4386868435C2776257435 @default.
- W4386868435 hasConceptScore W4386868435C2777212361 @default.
- W4386868435 hasConceptScore W4386868435C31258907 @default.
- W4386868435 hasConceptScore W4386868435C33923547 @default.
- W4386868435 hasConceptScore W4386868435C41008148 @default.
- W4386868435 hasConceptScore W4386868435C58471807 @default.
- W4386868435 hasConceptScore W4386868435C75294576 @default.
- W4386868435 hasConceptScore W4386868435C81363708 @default.
- W4386868435 hasLocation W43868684351 @default.
- W4386868435 hasOpenAccess W4386868435 @default.
- W4386868435 hasPrimaryLocation W43868684351 @default.
- W4386868435 hasRelatedWork W1665455280 @default.
- W4386868435 hasRelatedWork W2767651786 @default.
- W4386868435 hasRelatedWork W2912288872 @default.
- W4386868435 hasRelatedWork W2963331533 @default.
- W4386868435 hasRelatedWork W3021503072 @default.
- W4386868435 hasRelatedWork W3176807344 @default.
- W4386868435 hasRelatedWork W4206583062 @default.
- W4386868435 hasRelatedWork W4382680690 @default.
- W4386868435 hasRelatedWork W4386229954 @default.
- W4386868435 hasRelatedWork W4386911245 @default.
- W4386868435 isParatext "false" @default.
- W4386868435 isRetracted "false" @default.
- W4386868435 workType "article" @default.