Is your style transfer doing anything useful? an investigation into hippocampus segmentation and the role of preprocessing
Brain atrophy assessment in MRI, particularly of the hippocampus, is commonly used to support diagnosis and monitoring of dementia. Consequently, there is a demand for accurate automated hippocampus quantification. Most existing segmentation methods have been developed and validated on research data...
| Main Authors: | , , , , , , , , |
|---|---|
| Format: | Internet publication |
| Language: | English |
| Published: |
medRxiv
2024
|
| _version_ | 1826314746796703744 |
|---|---|
| author | Kalabizadeh, H Griffanti, L Yeung, P-H Voets, N Gillis, G Mackay, C Namburete, AI Dinsdale, NK Kamnitsas, K |
| author_facet | Kalabizadeh, H Griffanti, L Yeung, P-H Voets, N Gillis, G Mackay, C Namburete, AI Dinsdale, NK Kamnitsas, K |
| author_sort | Kalabizadeh, H |
| collection | OXFORD |
| description | Brain atrophy assessment in MRI, particularly of the hippocampus, is commonly used to support diagnosis and monitoring of dementia. Consequently, there is a demand for accurate automated hippocampus quantification. Most existing segmentation methods have been developed and validated on research datasets and, therefore, may not be appropriate for clinical MR images and populations, leading to potential gaps between dementia research and clinical practice. In this study, we investigated the performance of segmentation models trained on research data that were style-transferred to resemble clinical scans. Our results highlighted the importance of intensity normalisation methods in MRI segmentation, and their relation to domain shift and style-transfer. We found that whilst normalising intensity based on min and max values, commonly used in generative MR harmonisation methods, may create a need for style transfer, Z-score normalisation effectively maintains style consistency, and optimises performance. Moreover, we show for our datasets spatial augmentations are more beneficial than style harmonisation. Thus, emphasising robust normalisation techniques and spatial augmentation significantly improves MRI hippocampus segmentation. |
| first_indexed | 2024-12-09T03:10:27Z |
| format | Internet publication |
| id | oxford-uuid:cf941f69-0377-4d21-8f8c-3f8544b41281 |
| institution | University of Oxford |
| language | English |
| last_indexed | 2024-12-09T03:10:27Z |
| publishDate | 2024 |
| publisher | medRxiv |
| record_format | dspace |
| spelling | oxford-uuid:cf941f69-0377-4d21-8f8c-3f8544b412812024-10-03T12:29:49ZIs your style transfer doing anything useful? an investigation into hippocampus segmentation and the role of preprocessingInternet publicationhttp://purl.org/coar/resource_type/c_7ad9uuid:cf941f69-0377-4d21-8f8c-3f8544b41281EnglishSymplectic ElementsmedRxiv2024Kalabizadeh, HGriffanti, LYeung, P-HVoets, NGillis, GMackay, CNamburete, AIDinsdale, NKKamnitsas, KBrain atrophy assessment in MRI, particularly of the hippocampus, is commonly used to support diagnosis and monitoring of dementia. Consequently, there is a demand for accurate automated hippocampus quantification. Most existing segmentation methods have been developed and validated on research datasets and, therefore, may not be appropriate for clinical MR images and populations, leading to potential gaps between dementia research and clinical practice. In this study, we investigated the performance of segmentation models trained on research data that were style-transferred to resemble clinical scans. Our results highlighted the importance of intensity normalisation methods in MRI segmentation, and their relation to domain shift and style-transfer. We found that whilst normalising intensity based on min and max values, commonly used in generative MR harmonisation methods, may create a need for style transfer, Z-score normalisation effectively maintains style consistency, and optimises performance. Moreover, we show for our datasets spatial augmentations are more beneficial than style harmonisation. Thus, emphasising robust normalisation techniques and spatial augmentation significantly improves MRI hippocampus segmentation. |
| spellingShingle | Kalabizadeh, H Griffanti, L Yeung, P-H Voets, N Gillis, G Mackay, C Namburete, AI Dinsdale, NK Kamnitsas, K Is your style transfer doing anything useful? an investigation into hippocampus segmentation and the role of preprocessing |
| title | Is your style transfer doing anything useful? an investigation into hippocampus segmentation and the role of preprocessing |
| title_full | Is your style transfer doing anything useful? an investigation into hippocampus segmentation and the role of preprocessing |
| title_fullStr | Is your style transfer doing anything useful? an investigation into hippocampus segmentation and the role of preprocessing |
| title_full_unstemmed | Is your style transfer doing anything useful? an investigation into hippocampus segmentation and the role of preprocessing |
| title_short | Is your style transfer doing anything useful? an investigation into hippocampus segmentation and the role of preprocessing |
| title_sort | is your style transfer doing anything useful an investigation into hippocampus segmentation and the role of preprocessing |
| work_keys_str_mv | AT kalabizadehh isyourstyletransferdoinganythingusefulaninvestigationintohippocampussegmentationandtheroleofpreprocessing AT griffantil isyourstyletransferdoinganythingusefulaninvestigationintohippocampussegmentationandtheroleofpreprocessing AT yeungph isyourstyletransferdoinganythingusefulaninvestigationintohippocampussegmentationandtheroleofpreprocessing AT voetsn isyourstyletransferdoinganythingusefulaninvestigationintohippocampussegmentationandtheroleofpreprocessing AT gillisg isyourstyletransferdoinganythingusefulaninvestigationintohippocampussegmentationandtheroleofpreprocessing AT mackayc isyourstyletransferdoinganythingusefulaninvestigationintohippocampussegmentationandtheroleofpreprocessing AT nambureteai isyourstyletransferdoinganythingusefulaninvestigationintohippocampussegmentationandtheroleofpreprocessing AT dinsdalenk isyourstyletransferdoinganythingusefulaninvestigationintohippocampussegmentationandtheroleofpreprocessing AT kamnitsask isyourstyletransferdoinganythingusefulaninvestigationintohippocampussegmentationandtheroleofpreprocessing |