FRINK Linked Data Fragments server

Matches in SemOpenAlex for { <https://semopenalex.org/work/W4298111569> ?p ?o ?g. }

• W4298111569 endingPage "106157" @default.
• W4298111569 startingPage "106157" @default.
• W4298111569 abstract "Medical image segmentation is an important field in medical image analysis and a vital part of computer-aided diagnosis. Due to the challenges in acquiring image annotations, semi-supervised learning has attracted high attention in medical image segmentation. Despite their impressive performance, most existing semi-supervised approaches lack attention to ambiguous regions (e.g., some edges or corners around the organs). To achieve better performance, we propose a novel semi-supervised method called Adaptive Loss Balancing based on Homoscedastic Uncertainty in Multi-task Medical Image Segmentation Network (AHU-MultiNet). This model contains the main task for segmentation, one auxiliary task for signed distance, and another auxiliary task for contour detection. Our multi-task approach can effectively and sufficiently extract the semantic information of medical images by auxiliary tasks. Simultaneously, we introduce an inter-task consistency to explore the underlying information of the images and regularize the predictions in the right direction. More importantly, we notice and analyze that searching an optimal weighting manually to balance each task is a difficult and time-consuming process. Therefore, we introduce an adaptive loss balancing strategy based on homoscedastic uncertainty. Experimental results show that the two auxiliary tasks explicitly enforce shape-priors on the segmentation output to further generate more accurate masks under the adaptive loss balancing strategy. On several standard benchmarks, the 2018 Atrial Segmentation Challenge and the 2017 Liver Tumor Segmentation Challenge, our proposed method achieves improvements and outperforms the new state-of-the-art in semi-supervised learning. • We propose a novel multi-task framework for semi-supervised medical image segmentation. Three tasks use the information contained in their own domain to complement each other. We show that the inclusion of auxiliary tasks alongside the segmentation task improves performance. • We design a weight assignment using homoscedastic uncertainties confidence to dynamically tune the weights of multi-task losses. The task-specific homoscedastic uncertainties as noise parameters are updated together with model parameters during training. Compared with uniform weights, dynamic weightings balance the tasks optimally. • Our method obtain superior segmentation performance compared with existing semi-supervised methods on two common medical datasets." @default.
• W4298111569 created "2022-10-01" @default.
• W4298111569 creator A5008405013 @default.
• W4298111569 creator A5012772285 @default.
• W4298111569 creator A5015983532 @default.
• W4298111569 creator A5027627793 @default.
• W4298111569 creator A5033967745 @default.
• W4298111569 creator A5038805449 @default.
• W4298111569 date "2022-11-01" @default.
• W4298111569 modified "2023-10-14" @default.
• W4298111569 title "AHU-MultiNet: Adaptive loss balancing based on homoscedastic uncertainty in multi-task medical image segmentation network" @default.
• W4298111569 cites W2592929672 @default.
• W4298111569 cites W2962914239 @default.
• W4298111569 cites W3005650525 @default.
• W4298111569 cites W3024821730 @default.
• W4298111569 cites W3035665735 @default.
• W4298111569 cites W3040298941 @default.
• W4298111569 cites W3082954719 @default.
• W4298111569 cites W3098712157 @default.
• W4298111569 cites W3106904274 @default.
• W4298111569 cites W3196801498 @default.
• W4298111569 cites W3201787829 @default.
• W4298111569 cites W3206985657 @default.
• W4298111569 cites W4200120154 @default.
• W4298111569 cites W4210442314 @default.
• W4298111569 cites W4214935340 @default.
• W4298111569 cites W4280521522 @default.
• W4298111569 doi "https://doi.org/10.1016/j.compbiomed.2022.106157" @default.
• W4298111569 hasPublicationYear "2022" @default.
• W4298111569 type Work @default.
• W4298111569 citedByCount "2" @default.
• W4298111569 countsByYear W42981115692023 @default.
• W4298111569 crossrefType "journal-article" @default.
• W4298111569 hasAuthorship W4298111569A5008405013 @default.
• W4298111569 hasAuthorship W4298111569A5012772285 @default.
• W4298111569 hasAuthorship W4298111569A5015983532 @default.
• W4298111569 hasAuthorship W4298111569A5027627793 @default.
• W4298111569 hasAuthorship W4298111569A5033967745 @default.
• W4298111569 hasAuthorship W4298111569A5038805449 @default.
• W4298111569 hasConcept C101104100 @default.
• W4298111569 hasConcept C104409967 @default.
• W4298111569 hasConcept C115961682 @default.
• W4298111569 hasConcept C119857082 @default.
• W4298111569 hasConcept C124504099 @default.
• W4298111569 hasConcept C127413603 @default.
• W4298111569 hasConcept C154945302 @default.
• W4298111569 hasConcept C201995342 @default.
• W4298111569 hasConcept C2780451532 @default.
• W4298111569 hasConcept C31972630 @default.
• W4298111569 hasConcept C41008148 @default.
• W4298111569 hasConcept C89600930 @default.
• W4298111569 hasConceptScore W4298111569C101104100 @default.
• W4298111569 hasConceptScore W4298111569C104409967 @default.
• W4298111569 hasConceptScore W4298111569C115961682 @default.
• W4298111569 hasConceptScore W4298111569C119857082 @default.
• W4298111569 hasConceptScore W4298111569C124504099 @default.
• W4298111569 hasConceptScore W4298111569C127413603 @default.
• W4298111569 hasConceptScore W4298111569C154945302 @default.
• W4298111569 hasConceptScore W4298111569C201995342 @default.
• W4298111569 hasConceptScore W4298111569C2780451532 @default.
• W4298111569 hasConceptScore W4298111569C31972630 @default.
• W4298111569 hasConceptScore W4298111569C41008148 @default.
• W4298111569 hasConceptScore W4298111569C89600930 @default.
• W4298111569 hasFunder F4320321001 @default.
• W4298111569 hasLocation W42981115691 @default.
• W4298111569 hasOpenAccess W4298111569 @default.
• W4298111569 hasPrimaryLocation W42981115691 @default.
• W4298111569 hasRelatedWork W1507266234 @default.
• W4298111569 hasRelatedWork W1631910785 @default.
• W4298111569 hasRelatedWork W1669643531 @default.
• W4298111569 hasRelatedWork W1721780360 @default.
• W4298111569 hasRelatedWork W2110230079 @default.
• W4298111569 hasRelatedWork W2117664411 @default.
• W4298111569 hasRelatedWork W2117933325 @default.
• W4298111569 hasRelatedWork W2122581818 @default.
• W4298111569 hasRelatedWork W2159066190 @default.
• W4298111569 hasRelatedWork W2739874619 @default.
• W4298111569 hasVolume "150" @default.
• W4298111569 isParatext "false" @default.
• W4298111569 isRetracted "false" @default.
• W4298111569 workType "article" @default.