Matches in SemOpenAlex for { <https://semopenalex.org/work/W2914486560> ?p ?o ?g. }
Showing items 1 to 72 of
72
with 100 items per page.
- W2914486560 abstract "Introduction: Intracerebral Hemorrhage (ICH) accounts for 15% of strokes with a disproportionate impact on functional outcome and economic burden. In clinical trails, ICH volume on computed tomography (CT) is treated as both a criterion for inclusion and a radiographic outcome measure. Accurate assessment of ICH volume by manual segmentation is time consuming and clinical trials often use simple linear metrics such as ABC/2 despite their difficulty with large, irregular or lobar hematomas. Convolutional Neural Networks (CNNs) have shown to be adept at handling complex neuroimaging segmentation problems and have yet to be applied to an ICH clinical trial dataset. Methods: Initial pre-randomization CT scans from 112 patients with primary intracerebral hemorrhage from the MISTIE II trial dataset were manually segmented by two expert raters in OsiriX MD (Pixmeo) and divided into a training set (n = 72) and validation set (n = 40). The intracranial volume was extracted and data augmentation was applied to the training set using random axial rotation and elastic deformation. Each scan was then fed en bloc into a customized 3D CNN written in Tensorflow (Google) and trained over 50 data epochs in 24 hours. After standardized preprocessing, the validation dataset was fed into the network sequentially. Results: Individual segmentation predictions were generated in less than one second. Preliminary results, reported in median and interquartile range, include a Dice score of 0.90 (0.88 - 0.92). For ICH < 30cc (n = 15), 30 - 60cc (n = 19) and >60cc (n = 6), the Dice scores were 0.91 (0.87 - 0.93), 0.91 (0.87 - 0.92) and 0.89 (0.89 - 0.90) respectively. The median difference between predicted and actual volumes was 2.4cc (1.4 - 5.0cc) and the correlation between volumes was 0.97. Conclusion: Our results indicate that 3D CNNs can perform fast, accurate segmentation of ICH, including large hematomas. This model constitutes a potential tool for investigators requiring rapid analysis or working on large datasets." @default.
- W2914486560 created "2019-02-21" @default.
- W2914486560 creator A5005164866 @default.
- W2914486560 creator A5064742644 @default.
- W2914486560 creator A5066704933 @default.
- W2914486560 creator A5068628432 @default.
- W2914486560 creator A5087393296 @default.
- W2914486560 date "2019-02-01" @default.
- W2914486560 modified "2023-09-26" @default.
- W2914486560 title "Abstract WMP13: 3D Convolutional Neural Network Segmentation of Intracerebral Hemorrhage on CT" @default.
- W2914486560 doi "https://doi.org/10.1161/str.50.suppl_1.wmp13" @default.
- W2914486560 hasPublicationYear "2019" @default.
- W2914486560 type Work @default.
- W2914486560 sameAs 2914486560 @default.
- W2914486560 citedByCount "0" @default.
- W2914486560 crossrefType "journal-article" @default.
- W2914486560 hasAuthorship W2914486560A5005164866 @default.
- W2914486560 hasAuthorship W2914486560A5064742644 @default.
- W2914486560 hasAuthorship W2914486560A5066704933 @default.
- W2914486560 hasAuthorship W2914486560A5068628432 @default.
- W2914486560 hasAuthorship W2914486560A5087393296 @default.
- W2914486560 hasConcept C118552586 @default.
- W2914486560 hasConcept C119060515 @default.
- W2914486560 hasConcept C126838900 @default.
- W2914486560 hasConcept C141071460 @default.
- W2914486560 hasConcept C154945302 @default.
- W2914486560 hasConcept C168563851 @default.
- W2914486560 hasConcept C204243189 @default.
- W2914486560 hasConcept C2777094939 @default.
- W2914486560 hasConcept C2777736543 @default.
- W2914486560 hasConcept C2989005 @default.
- W2914486560 hasConcept C41008148 @default.
- W2914486560 hasConcept C58489278 @default.
- W2914486560 hasConcept C58693492 @default.
- W2914486560 hasConcept C71924100 @default.
- W2914486560 hasConcept C81363708 @default.
- W2914486560 hasConcept C89600930 @default.
- W2914486560 hasConceptScore W2914486560C118552586 @default.
- W2914486560 hasConceptScore W2914486560C119060515 @default.
- W2914486560 hasConceptScore W2914486560C126838900 @default.
- W2914486560 hasConceptScore W2914486560C141071460 @default.
- W2914486560 hasConceptScore W2914486560C154945302 @default.
- W2914486560 hasConceptScore W2914486560C168563851 @default.
- W2914486560 hasConceptScore W2914486560C204243189 @default.
- W2914486560 hasConceptScore W2914486560C2777094939 @default.
- W2914486560 hasConceptScore W2914486560C2777736543 @default.
- W2914486560 hasConceptScore W2914486560C2989005 @default.
- W2914486560 hasConceptScore W2914486560C41008148 @default.
- W2914486560 hasConceptScore W2914486560C58489278 @default.
- W2914486560 hasConceptScore W2914486560C58693492 @default.
- W2914486560 hasConceptScore W2914486560C71924100 @default.
- W2914486560 hasConceptScore W2914486560C81363708 @default.
- W2914486560 hasConceptScore W2914486560C89600930 @default.
- W2914486560 hasIssue "Suppl_1" @default.
- W2914486560 hasLocation W29144865601 @default.
- W2914486560 hasOpenAccess W2914486560 @default.
- W2914486560 hasPrimaryLocation W29144865601 @default.
- W2914486560 hasRelatedWork W2673946014 @default.
- W2914486560 hasRelatedWork W2795329967 @default.
- W2914486560 hasRelatedWork W2899211859 @default.
- W2914486560 hasRelatedWork W2963940192 @default.
- W2914486560 hasRelatedWork W2994948129 @default.
- W2914486560 hasRelatedWork W3102253946 @default.
- W2914486560 hasRelatedWork W3144574764 @default.
- W2914486560 hasRelatedWork W3148584990 @default.
- W2914486560 hasRelatedWork W4200528772 @default.
- W2914486560 hasRelatedWork W4293211451 @default.
- W2914486560 hasVolume "50" @default.
- W2914486560 isParatext "false" @default.
- W2914486560 isRetracted "false" @default.
- W2914486560 magId "2914486560" @default.
- W2914486560 workType "article" @default.