Convolutional Neural Networks for Classifying Laterality of Vestibular Schwannomas on Single MRI Slices—A Feasibility Study
<b>Introduction</b>: Many proposed algorithms for tumor detection rely on 2.5/3D convolutional neural networks (CNNs) and the input of segmentations for training. The purpose of this study is therefore to assess the performance of tumor detection on single MRI slices containing vestibula...
| Main Authors: | , , , , , , , , , , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
MDPI AG
2021-09-01
|
| Series: | Diagnostics |
| Subjects: | |
| Online Access: | https://www.mdpi.com/2075-4418/11/9/1676 |
| _version_ | 1797519609425821696 |
|---|---|
| author | Philipp Sager Lukas Näf Erwin Vu Tim Fischer Paul M. Putora Felix Ehret Christoph Fürweger Christina Schröder Robert Förster Daniel R. Zwahlen Alexander Muacevic Paul Windisch |
| author_facet | Philipp Sager Lukas Näf Erwin Vu Tim Fischer Paul M. Putora Felix Ehret Christoph Fürweger Christina Schröder Robert Förster Daniel R. Zwahlen Alexander Muacevic Paul Windisch |
| author_sort | Philipp Sager |
| collection | DOAJ |
| description | <b>Introduction</b>: Many proposed algorithms for tumor detection rely on 2.5/3D convolutional neural networks (CNNs) and the input of segmentations for training. The purpose of this study is therefore to assess the performance of tumor detection on single MRI slices containing vestibular schwannomas (VS) as a computationally inexpensive alternative that does not require the creation of segmentations. <b>Methods</b>: A total of 2992 T1-weighted contrast-enhanced axial slices containing VS from the MRIs of 633 patients were labeled according to tumor location, of which 2538 slices from 539 patients were used for training a CNN (ResNet-34) to classify them according to the side of the tumor as a surrogate for detection and 454 slices from 94 patients were used for internal validation. The model was then externally validated on contrast-enhanced and non-contrast-enhanced slices from a different institution. Categorical accuracy was noted, and the results of the predictions for the validation set are provided with confusion matrices. <b>Results</b>: The model achieved an accuracy of 0.928 (95% CI: 0.869–0.987) on contrast-enhanced slices and 0.795 (95% CI: 0.702–0.888) on non-contrast-enhanced slices from the external validation cohorts. The implementation of Gradient-weighted Class Activation Mapping (Grad-CAM) revealed that the focus of the model was not limited to the contrast-enhancing tumor but to a larger area of the cerebellum and the cerebellopontine angle. <b>Conclusions</b>: Single-slice predictions might constitute a computationally inexpensive alternative to training 2.5/3D-CNNs for certain detection tasks in medical imaging even without the use of segmentations. Head-to-head comparisons between 2D and more sophisticated architectures could help to determine the difference in accuracy, especially for more difficult tasks. |
| first_indexed | 2024-03-10T07:45:13Z |
| format | Article |
| id | doaj.art-29c3347e1f164b03b0e717baf1786d7d |
| institution | Directory Open Access Journal |
| issn | 2075-4418 |
| language | English |
| last_indexed | 2024-03-10T07:45:13Z |
| publishDate | 2021-09-01 |
| publisher | MDPI AG |
| record_format | Article |
| series | Diagnostics |
| spelling | doaj.art-29c3347e1f164b03b0e717baf1786d7d2023-11-22T12:40:42ZengMDPI AGDiagnostics2075-44182021-09-01119167610.3390/diagnostics11091676Convolutional Neural Networks for Classifying Laterality of Vestibular Schwannomas on Single MRI Slices—A Feasibility StudyPhilipp Sager0Lukas Näf1Erwin Vu2Tim Fischer3Paul M. Putora4Felix Ehret5Christoph Fürweger6Christina Schröder7Robert Förster8Daniel R. Zwahlen9Alexander Muacevic10Paul Windisch11Department of Radiation Oncology, Kantonsspital Winterthur, 8400 Winterthur, SwitzerlandDepartment of Radiology, Kantonsspital St. Gallen, 9007 St. Gallen, SwitzerlandDepartment of Radiation Oncology, Kantonsspital St. Gallen, 9007 St. Gallen, SwitzerlandDepartment of Radiology, Kantonsspital St. Gallen, 9007 St. Gallen, SwitzerlandDepartment of Radiation Oncology, Kantonsspital St. Gallen, 9007 St. Gallen, SwitzerlandCharité—Universitätsmedizin Berlin, Corporate Member of Freie Universität Berlin and Humboldt-Universität zu Berlin, Department of Radiation Oncology, 13353 Berlin, GermanyEuropean Cyberknife Center, 81377 Munich, GermanyDepartment of Radiation Oncology, Kantonsspital Winterthur, 8400 Winterthur, SwitzerlandDepartment of Radiation Oncology, Kantonsspital Winterthur, 8400 Winterthur, SwitzerlandDepartment of Radiation Oncology, Kantonsspital Winterthur, 8400 Winterthur, SwitzerlandEuropean Cyberknife Center, 81377 Munich, GermanyDepartment of Radiation Oncology, Kantonsspital Winterthur, 8400 Winterthur, Switzerland<b>Introduction</b>: Many proposed algorithms for tumor detection rely on 2.5/3D convolutional neural networks (CNNs) and the input of segmentations for training. The purpose of this study is therefore to assess the performance of tumor detection on single MRI slices containing vestibular schwannomas (VS) as a computationally inexpensive alternative that does not require the creation of segmentations. <b>Methods</b>: A total of 2992 T1-weighted contrast-enhanced axial slices containing VS from the MRIs of 633 patients were labeled according to tumor location, of which 2538 slices from 539 patients were used for training a CNN (ResNet-34) to classify them according to the side of the tumor as a surrogate for detection and 454 slices from 94 patients were used for internal validation. The model was then externally validated on contrast-enhanced and non-contrast-enhanced slices from a different institution. Categorical accuracy was noted, and the results of the predictions for the validation set are provided with confusion matrices. <b>Results</b>: The model achieved an accuracy of 0.928 (95% CI: 0.869–0.987) on contrast-enhanced slices and 0.795 (95% CI: 0.702–0.888) on non-contrast-enhanced slices from the external validation cohorts. The implementation of Gradient-weighted Class Activation Mapping (Grad-CAM) revealed that the focus of the model was not limited to the contrast-enhancing tumor but to a larger area of the cerebellum and the cerebellopontine angle. <b>Conclusions</b>: Single-slice predictions might constitute a computationally inexpensive alternative to training 2.5/3D-CNNs for certain detection tasks in medical imaging even without the use of segmentations. Head-to-head comparisons between 2D and more sophisticated architectures could help to determine the difference in accuracy, especially for more difficult tasks.https://www.mdpi.com/2075-4418/11/9/1676artificial intelligencedeep learningmachine learningvestibularschwannomaneuro-oncology |
| spellingShingle | Philipp Sager Lukas Näf Erwin Vu Tim Fischer Paul M. Putora Felix Ehret Christoph Fürweger Christina Schröder Robert Förster Daniel R. Zwahlen Alexander Muacevic Paul Windisch Convolutional Neural Networks for Classifying Laterality of Vestibular Schwannomas on Single MRI Slices—A Feasibility Study Diagnostics artificial intelligence deep learning machine learning vestibular schwannoma neuro-oncology |
| title | Convolutional Neural Networks for Classifying Laterality of Vestibular Schwannomas on Single MRI Slices—A Feasibility Study |
| title_full | Convolutional Neural Networks for Classifying Laterality of Vestibular Schwannomas on Single MRI Slices—A Feasibility Study |
| title_fullStr | Convolutional Neural Networks for Classifying Laterality of Vestibular Schwannomas on Single MRI Slices—A Feasibility Study |
| title_full_unstemmed | Convolutional Neural Networks for Classifying Laterality of Vestibular Schwannomas on Single MRI Slices—A Feasibility Study |
| title_short | Convolutional Neural Networks for Classifying Laterality of Vestibular Schwannomas on Single MRI Slices—A Feasibility Study |
| title_sort | convolutional neural networks for classifying laterality of vestibular schwannomas on single mri slices a feasibility study |
| topic | artificial intelligence deep learning machine learning vestibular schwannoma neuro-oncology |
| url | https://www.mdpi.com/2075-4418/11/9/1676 |
| work_keys_str_mv | AT philippsager convolutionalneuralnetworksforclassifyinglateralityofvestibularschwannomasonsinglemrislicesafeasibilitystudy AT lukasnaf convolutionalneuralnetworksforclassifyinglateralityofvestibularschwannomasonsinglemrislicesafeasibilitystudy AT erwinvu convolutionalneuralnetworksforclassifyinglateralityofvestibularschwannomasonsinglemrislicesafeasibilitystudy AT timfischer convolutionalneuralnetworksforclassifyinglateralityofvestibularschwannomasonsinglemrislicesafeasibilitystudy AT paulmputora convolutionalneuralnetworksforclassifyinglateralityofvestibularschwannomasonsinglemrislicesafeasibilitystudy AT felixehret convolutionalneuralnetworksforclassifyinglateralityofvestibularschwannomasonsinglemrislicesafeasibilitystudy AT christophfurweger convolutionalneuralnetworksforclassifyinglateralityofvestibularschwannomasonsinglemrislicesafeasibilitystudy AT christinaschroder convolutionalneuralnetworksforclassifyinglateralityofvestibularschwannomasonsinglemrislicesafeasibilitystudy AT robertforster convolutionalneuralnetworksforclassifyinglateralityofvestibularschwannomasonsinglemrislicesafeasibilitystudy AT danielrzwahlen convolutionalneuralnetworksforclassifyinglateralityofvestibularschwannomasonsinglemrislicesafeasibilitystudy AT alexandermuacevic convolutionalneuralnetworksforclassifyinglateralityofvestibularschwannomasonsinglemrislicesafeasibilitystudy AT paulwindisch convolutionalneuralnetworksforclassifyinglateralityofvestibularschwannomasonsinglemrislicesafeasibilitystudy |