Evaluation of the HD-GLIO Deep Learning Algorithm for Brain Tumour Segmentation on Postoperative MRI
In the context of brain tumour response assessment, deep learning-based three-dimensional (3D) tumour segmentation has shown potential to enter the routine radiological workflow. The purpose of the present study was to perform an external evaluation of a state-of-the-art deep learning 3D brain tumou...
| Main Authors: | , , , , , , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
MDPI AG
2023-01-01
|
| Series: | Diagnostics |
| Subjects: | |
| Online Access: | https://www.mdpi.com/2075-4418/13/3/363 |
| _version_ | 1797624909328809984 |
|---|---|
| author | Peter Jagd Sørensen Jonathan Frederik Carlsen Vibeke Andrée Larsen Flemming Littrup Andersen Claes Nøhr Ladefoged Michael Bachmann Nielsen Hans Skovgaard Poulsen Adam Espe Hansen |
| author_facet | Peter Jagd Sørensen Jonathan Frederik Carlsen Vibeke Andrée Larsen Flemming Littrup Andersen Claes Nøhr Ladefoged Michael Bachmann Nielsen Hans Skovgaard Poulsen Adam Espe Hansen |
| author_sort | Peter Jagd Sørensen |
| collection | DOAJ |
| description | In the context of brain tumour response assessment, deep learning-based three-dimensional (3D) tumour segmentation has shown potential to enter the routine radiological workflow. The purpose of the present study was to perform an external evaluation of a state-of-the-art deep learning 3D brain tumour segmentation algorithm (HD-GLIO) on an independent cohort of consecutive, post-operative patients. For 66 consecutive magnetic resonance imaging examinations, we compared delineations of contrast-enhancing (CE) tumour lesions and non-enhancing T2/FLAIR hyperintense abnormality (NE) lesions by the HD-GLIO algorithm and radiologists using Dice similarity coefficients (Dice). Volume agreement was assessed using concordance correlation coefficients (CCCs) and Bland–Altman plots. The algorithm performed very well regarding the segmentation of NE volumes (median Dice = 0.79) and CE tumour volumes larger than 1.0 cm<sup>3</sup> (median Dice = 0.86). If considering all cases with CE tumour lesions, the performance dropped significantly (median Dice = 0.40). Volume agreement was excellent with CCCs of 0.997 (CE tumour volumes) and 0.922 (NE volumes). The findings have implications for the application of the HD-GLIO algorithm in the routine radiological workflow where small contrast-enhancing tumours will constitute a considerable share of the follow-up cases. Our study underlines that independent validations on clinical datasets are key to asserting the robustness of deep learning algorithms. |
| first_indexed | 2024-03-11T09:49:19Z |
| format | Article |
| id | doaj.art-928982add86448a38e3ec258721352eb |
| institution | Directory Open Access Journal |
| issn | 2075-4418 |
| language | English |
| last_indexed | 2024-03-11T09:49:19Z |
| publishDate | 2023-01-01 |
| publisher | MDPI AG |
| record_format | Article |
| series | Diagnostics |
| spelling | doaj.art-928982add86448a38e3ec258721352eb2023-11-16T16:23:38ZengMDPI AGDiagnostics2075-44182023-01-0113336310.3390/diagnostics13030363Evaluation of the HD-GLIO Deep Learning Algorithm for Brain Tumour Segmentation on Postoperative MRIPeter Jagd Sørensen0Jonathan Frederik Carlsen1Vibeke Andrée Larsen2Flemming Littrup Andersen3Claes Nøhr Ladefoged4Michael Bachmann Nielsen5Hans Skovgaard Poulsen6Adam Espe Hansen7Department of Radiology, Centre of Diagnostic Investigation, Copenhagen University Hospital—Rigshospitalet, 2100 Copenhagen, DenmarkDepartment of Radiology, Centre of Diagnostic Investigation, Copenhagen University Hospital—Rigshospitalet, 2100 Copenhagen, DenmarkDepartment of Radiology, Centre of Diagnostic Investigation, Copenhagen University Hospital—Rigshospitalet, 2100 Copenhagen, DenmarkDepartment of Clinical Medicine, University of Copenhagen, 2100 Copenhagen, DenmarkDepartment of Clinical Physiology and Nuclear Medicine, Centre of Diagnostic Investigation, Copenhagen University Hospital—Rigshospitalet, 2100 Copenhagen, DenmarkDepartment of Radiology, Centre of Diagnostic Investigation, Copenhagen University Hospital—Rigshospitalet, 2100 Copenhagen, DenmarkThe DCCC Brain Tumor Center, 2100 Copenhagen, DenmarkDepartment of Radiology, Centre of Diagnostic Investigation, Copenhagen University Hospital—Rigshospitalet, 2100 Copenhagen, DenmarkIn the context of brain tumour response assessment, deep learning-based three-dimensional (3D) tumour segmentation has shown potential to enter the routine radiological workflow. The purpose of the present study was to perform an external evaluation of a state-of-the-art deep learning 3D brain tumour segmentation algorithm (HD-GLIO) on an independent cohort of consecutive, post-operative patients. For 66 consecutive magnetic resonance imaging examinations, we compared delineations of contrast-enhancing (CE) tumour lesions and non-enhancing T2/FLAIR hyperintense abnormality (NE) lesions by the HD-GLIO algorithm and radiologists using Dice similarity coefficients (Dice). Volume agreement was assessed using concordance correlation coefficients (CCCs) and Bland–Altman plots. The algorithm performed very well regarding the segmentation of NE volumes (median Dice = 0.79) and CE tumour volumes larger than 1.0 cm<sup>3</sup> (median Dice = 0.86). If considering all cases with CE tumour lesions, the performance dropped significantly (median Dice = 0.40). Volume agreement was excellent with CCCs of 0.997 (CE tumour volumes) and 0.922 (NE volumes). The findings have implications for the application of the HD-GLIO algorithm in the routine radiological workflow where small contrast-enhancing tumours will constitute a considerable share of the follow-up cases. Our study underlines that independent validations on clinical datasets are key to asserting the robustness of deep learning algorithms.https://www.mdpi.com/2075-4418/13/3/363brain tumour segmentationtreatment monitoringroutinepostoperativeautomaticdeep learning algorithm |
| spellingShingle | Peter Jagd Sørensen Jonathan Frederik Carlsen Vibeke Andrée Larsen Flemming Littrup Andersen Claes Nøhr Ladefoged Michael Bachmann Nielsen Hans Skovgaard Poulsen Adam Espe Hansen Evaluation of the HD-GLIO Deep Learning Algorithm for Brain Tumour Segmentation on Postoperative MRI Diagnostics brain tumour segmentation treatment monitoring routine postoperative automatic deep learning algorithm |
| title | Evaluation of the HD-GLIO Deep Learning Algorithm for Brain Tumour Segmentation on Postoperative MRI |
| title_full | Evaluation of the HD-GLIO Deep Learning Algorithm for Brain Tumour Segmentation on Postoperative MRI |
| title_fullStr | Evaluation of the HD-GLIO Deep Learning Algorithm for Brain Tumour Segmentation on Postoperative MRI |
| title_full_unstemmed | Evaluation of the HD-GLIO Deep Learning Algorithm for Brain Tumour Segmentation on Postoperative MRI |
| title_short | Evaluation of the HD-GLIO Deep Learning Algorithm for Brain Tumour Segmentation on Postoperative MRI |
| title_sort | evaluation of the hd glio deep learning algorithm for brain tumour segmentation on postoperative mri |
| topic | brain tumour segmentation treatment monitoring routine postoperative automatic deep learning algorithm |
| url | https://www.mdpi.com/2075-4418/13/3/363 |
| work_keys_str_mv | AT peterjagdsørensen evaluationofthehdgliodeeplearningalgorithmforbraintumoursegmentationonpostoperativemri AT jonathanfrederikcarlsen evaluationofthehdgliodeeplearningalgorithmforbraintumoursegmentationonpostoperativemri AT vibekeandreelarsen evaluationofthehdgliodeeplearningalgorithmforbraintumoursegmentationonpostoperativemri AT flemminglittrupandersen evaluationofthehdgliodeeplearningalgorithmforbraintumoursegmentationonpostoperativemri AT claesnøhrladefoged evaluationofthehdgliodeeplearningalgorithmforbraintumoursegmentationonpostoperativemri AT michaelbachmannnielsen evaluationofthehdgliodeeplearningalgorithmforbraintumoursegmentationonpostoperativemri AT hansskovgaardpoulsen evaluationofthehdgliodeeplearningalgorithmforbraintumoursegmentationonpostoperativemri AT adamespehansen evaluationofthehdgliodeeplearningalgorithmforbraintumoursegmentationonpostoperativemri |