BucketAugment: Reinforced Domain Generalisation in Abdominal CT Segmentation
<italic>Goal:</italic> In recent years, deep neural networks have consistently outperformed previously proposed methods in the domain of medical segmentation. However, due to their nature, these networks often struggle to delineate desired structures in data that fall outside their train...
Saved in:
| Main Authors: | , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
IEEE
2024-01-01
|
| Series: | IEEE Open Journal of Engineering in Medicine and Biology |
| Subjects: | |
| Online Access: | https://ieeexplore.ieee.org/document/10521822/ |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| _version_ | 1832582292483080192 |
|---|---|
| author | David Jozef Hresko Peter Drotar |
| author_facet | David Jozef Hresko Peter Drotar |
| author_sort | David Jozef Hresko |
| collection | DOAJ |
| description | <italic>Goal:</italic> In recent years, deep neural networks have consistently outperformed previously proposed methods in the domain of medical segmentation. However, due to their nature, these networks often struggle to delineate desired structures in data that fall outside their training distribution. The goal of this study is to address the challenges associated with domain generalization in CT segmentation by introducing a novel method called BucketAugment for deep neural networks. <italic>Methods:</italic> BucketAugment leverages principles from the Q-learning algorithm and employs validation loss to search for an optimal policy within a search space comprised of distributed stacks of 3D volumetric augmentations, termed ‘buckets.’ These buckets have tunable parameters and can be seamlessly integrated into existing neural network architectures, offering flexibility for customization. <italic>Results:</italic> In our experiments, we focus on segmenting kidney and liver structures across three distinct medical datasets, each containing CT scans of the abdominal region collected from various clinical institutions and scanner vendors. Our results indicate that BucketAugment significantly enhances domain generalization across diverse medical datasets, requiring only minimal modifications to existing network architectures. <italic>Conclusions:</italic> The introduction of BucketAugment provides a promising solution to the challenges of domain generalization in CT segmentation. By leveraging Q-learning principles and distributed stacks of 3D augmentations, this method improves the performance of deep neural networks on medical segmentation tasks, demonstrating its potential to enhance the applicability of such models across different datasets and clinical scenarios. |
| format | Article |
| id | doaj-art-2c6fdacf940f4675b2c09d6d186b957f |
| institution | Kabale University |
| issn | 2644-1276 |
| language | English |
| publishDate | 2024-01-01 |
| publisher | IEEE |
| record_format | Article |
| series | IEEE Open Journal of Engineering in Medicine and Biology |
| spelling | doaj-art-2c6fdacf940f4675b2c09d6d186b957f2025-01-30T00:03:42ZengIEEEIEEE Open Journal of Engineering in Medicine and Biology2644-12762024-01-01535336110.1109/OJEMB.2024.339762310521822BucketAugment: Reinforced Domain Generalisation in Abdominal CT SegmentationDavid Jozef Hresko0https://orcid.org/0000-0002-4887-3412Peter Drotar1https://orcid.org/0000-0002-6634-4696Technical University of Kosice, Kosice, SlovakiaTechnical University of Kosice, Kosice, Slovakia<italic>Goal:</italic> In recent years, deep neural networks have consistently outperformed previously proposed methods in the domain of medical segmentation. However, due to their nature, these networks often struggle to delineate desired structures in data that fall outside their training distribution. The goal of this study is to address the challenges associated with domain generalization in CT segmentation by introducing a novel method called BucketAugment for deep neural networks. <italic>Methods:</italic> BucketAugment leverages principles from the Q-learning algorithm and employs validation loss to search for an optimal policy within a search space comprised of distributed stacks of 3D volumetric augmentations, termed ‘buckets.’ These buckets have tunable parameters and can be seamlessly integrated into existing neural network architectures, offering flexibility for customization. <italic>Results:</italic> In our experiments, we focus on segmenting kidney and liver structures across three distinct medical datasets, each containing CT scans of the abdominal region collected from various clinical institutions and scanner vendors. Our results indicate that BucketAugment significantly enhances domain generalization across diverse medical datasets, requiring only minimal modifications to existing network architectures. <italic>Conclusions:</italic> The introduction of BucketAugment provides a promising solution to the challenges of domain generalization in CT segmentation. By leveraging Q-learning principles and distributed stacks of 3D augmentations, this method improves the performance of deep neural networks on medical segmentation tasks, demonstrating its potential to enhance the applicability of such models across different datasets and clinical scenarios.https://ieeexplore.ieee.org/document/10521822/Medical image segmentationimage augmentationdomain generalisationabdominal CTreinforcement learning |
| spellingShingle | David Jozef Hresko Peter Drotar BucketAugment: Reinforced Domain Generalisation in Abdominal CT Segmentation IEEE Open Journal of Engineering in Medicine and Biology Medical image segmentation image augmentation domain generalisation abdominal CT reinforcement learning |
| title | BucketAugment: Reinforced Domain Generalisation in Abdominal CT Segmentation |
| title_full | BucketAugment: Reinforced Domain Generalisation in Abdominal CT Segmentation |
| title_fullStr | BucketAugment: Reinforced Domain Generalisation in Abdominal CT Segmentation |
| title_full_unstemmed | BucketAugment: Reinforced Domain Generalisation in Abdominal CT Segmentation |
| title_short | BucketAugment: Reinforced Domain Generalisation in Abdominal CT Segmentation |
| title_sort | bucketaugment reinforced domain generalisation in abdominal ct segmentation |
| topic | Medical image segmentation image augmentation domain generalisation abdominal CT reinforcement learning |
| url | https://ieeexplore.ieee.org/document/10521822/ |
| work_keys_str_mv | AT davidjozefhresko bucketaugmentreinforceddomaingeneralisationinabdominalctsegmentation AT peterdrotar bucketaugmentreinforceddomaingeneralisationinabdominalctsegmentation |