FRINK Linked Data Fragments server

Matches in SemOpenAlex for { <https://semopenalex.org/work/W3165085080> ?p ?o ?g. }

• W3165085080 endingPage "1431" @default.
• W3165085080 startingPage "1431" @default.
• W3165085080 abstract "1431 Objectives: Accurate delineation of the gross tumor volume (GTV) is a key step for accurate and effective management of head and neck cancer patients. Manual segmentation of PET images for delineation of GTV or calculation of metabolic tumor volume (MTV) and total lesion glycolysis (TLG) is time consuming and prone to intra- and inter-observer variability. The availability of a fast and robust fully automated PET image segmentation algorithm is highly desired. The aim of the current study is to develop a fully automated image segmentation algorithm of head and neck malignant lesions from PET and CT images. Methods: We included 161 clinical PET/CT images of head and neck cancer patients. Eighty percent of the subjects were used for training whereas 20% were kept for external validation. The GTVs were manually delineated simultaneously on PET and CT images. The intensities of CT images in Hounsfield units and PET images converted to standardized uptake value (SUV) units were linearly normalized with respect to the maximum value of each dataset. PET and CT images were fused using wavelet fusion (WF), guided filtering-based fusion (GFF), and latent low-rank representation fusion (LLRR) algorithms. We implemented a modified multichannel input 3D U-NET for fully automatic segmentation. Standard image segmentation metrics, including Dice similarity index (Dice) were used for performance assessment of the algorithms. The relative error (RE%) of PET quantitative parameters including SUVmax, MTV and TLG were calculated. Results: The Dice for PET and CT only images was 0.79±0.01 and 0.73±0.08, respectively. This increased to 0.81±0.06 for multichannel PET and CT images and to 0.83±0.07 for WF fused images. Using multi-channel PET+CT+WF as input resulted in a Dice of 0.85±0.01. All images resulted in relative errors less than 1% for SUVmax whereas the lowest relative error was achieved for PET+CT+WF for MTV and TLG estimations (less than 5%). We also reported the outlier cases which deep learning algorithms failed to properly segment the GTVs.Conclusion: Deep learning-based algorithms exhibited promising performance for fully automated GTV segmentation of head and neck PET/CT images. Using multi-channel PET/CT and fusion image as input to the deep neural network improved the accuracy of GTV segmentation of head and neck lesions. Outlier cases should be considered in clinical routine." @default.
• W3165085080 created "2021-06-07" @default.
• W3165085080 creator A5003078255 @default.
• W3165085080 creator A5007891293 @default.
• W3165085080 creator A5034625862 @default.
• W3165085080 creator A5039181443 @default.
• W3165085080 date "2021-05-01" @default.
• W3165085080 modified "2023-09-23" @default.
• W3165085080 title "Fully Automated Head and Neck Malignant Lesions Segmentation using Multimodality PET/CT imaging and A Deep Convolutional Network" @default.
• W3165085080 hasPublicationYear "2021" @default.
• W3165085080 type Work @default.
• W3165085080 sameAs 3165085080 @default.
• W3165085080 citedByCount "0" @default.
• W3165085080 crossrefType "journal-article" @default.
• W3165085080 hasAuthorship W3165085080A5003078255 @default.
• W3165085080 hasAuthorship W3165085080A5007891293 @default.
• W3165085080 hasAuthorship W3165085080A5034625862 @default.
• W3165085080 hasAuthorship W3165085080A5039181443 @default.
• W3165085080 hasConcept C127077266 @default.
• W3165085080 hasConcept C154945302 @default.
• W3165085080 hasConcept C199374082 @default.
• W3165085080 hasConcept C22029948 @default.
• W3165085080 hasConcept C2524010 @default.
• W3165085080 hasConcept C2775842073 @default.
• W3165085080 hasConcept C2989005 @default.
• W3165085080 hasConcept C33923547 @default.
• W3165085080 hasConcept C41008148 @default.
• W3165085080 hasConcept C71924100 @default.
• W3165085080 hasConcept C81363708 @default.
• W3165085080 hasConcept C89600930 @default.
• W3165085080 hasConceptScore W3165085080C127077266 @default.
• W3165085080 hasConceptScore W3165085080C154945302 @default.
• W3165085080 hasConceptScore W3165085080C199374082 @default.
• W3165085080 hasConceptScore W3165085080C22029948 @default.
• W3165085080 hasConceptScore W3165085080C2524010 @default.
• W3165085080 hasConceptScore W3165085080C2775842073 @default.
• W3165085080 hasConceptScore W3165085080C2989005 @default.
• W3165085080 hasConceptScore W3165085080C33923547 @default.
• W3165085080 hasConceptScore W3165085080C41008148 @default.
• W3165085080 hasConceptScore W3165085080C71924100 @default.
• W3165085080 hasConceptScore W3165085080C81363708 @default.
• W3165085080 hasConceptScore W3165085080C89600930 @default.
• W3165085080 hasOpenAccess W3165085080 @default.
• W3165085080 hasRelatedWork W1965452697 @default.
• W3165085080 hasRelatedWork W2020530055 @default.
• W3165085080 hasRelatedWork W2027997001 @default.
• W3165085080 hasRelatedWork W2034946043 @default.
• W3165085080 hasRelatedWork W2059939049 @default.
• W3165085080 hasRelatedWork W2168057128 @default.
• W3165085080 hasRelatedWork W2169769956 @default.
• W3165085080 hasRelatedWork W2316075884 @default.
• W3165085080 hasRelatedWork W2351435554 @default.
• W3165085080 hasRelatedWork W2773257842 @default.
• W3165085080 hasRelatedWork W2898441006 @default.
• W3165085080 hasRelatedWork W2947071620 @default.
• W3165085080 hasRelatedWork W2947506679 @default.
• W3165085080 hasRelatedWork W2948770679 @default.
• W3165085080 hasRelatedWork W3012291095 @default.
• W3165085080 hasRelatedWork W3119000373 @default.
• W3165085080 hasRelatedWork W3164816110 @default.
• W3165085080 hasRelatedWork W3165545946 @default.
• W3165085080 hasRelatedWork W3194534285 @default.
• W3165085080 hasRelatedWork W2554092771 @default.
• W3165085080 hasVolume "62" @default.
• W3165085080 isParatext "false" @default.
• W3165085080 isRetracted "false" @default.
• W3165085080 magId "3165085080" @default.
• W3165085080 workType "article" @default.