Matches in SemOpenAlex for { <https://semopenalex.org/work/W4386362602> ?p ?o ?g. }
Showing items 1 to 84 of
84
with 100 items per page.
- W4386362602 abstract "Scarcity of labels for medical images is a significant barrier for training representation learning approaches based on deep neural networks. This limitation is also present when using imaging data collected during routine clinical care stored in picture archiving communication systems (PACS), as these data rarely have attached the high-quality labels required for medical image computing tasks. However, medical images extracted from PACS are commonly coupled with descriptive radiology reports that contain significant information and could be leveraged to pre-train imaging models, which could serve as starting points for further task-specific fine-tuning.In this work, we perform a head-to-head comparison of three different self-supervised strategies to pre-train the same imaging model on 3D brain computed tomography angiogram (CTA) images, with large vessel occlusion (LVO) detection as the downstream task. These strategies evaluate two natural language processing (NLP) approaches, one to extract 100 explicit radiology concepts (Rad-SpatialNet) and the other to create general-purpose radiology reports embeddings (Distil-BERT). In addition, we experiment with learning radiology concepts directly or by using a recent self-supervised learning approach (CLIP) that learns by ranking the distance between language and image vector embeddings. The LVO detection task was selected because it requires 3D imaging data, is clinically important, and requires the algorithm to learn outputs not explicitly stated in the radiology report.Pre-training was performed on an unlabeled dataset containing 1,542 3D CTA - reports pairs. The downstream task was tested on a labeled dataset of 402 subjects for LVO. We find that the pre-training performed with CLIP-based strategies improve the performance of the imaging model to detect LVO compared to a model trained only on the labeled data. The best performance was achieved by pre-training using the explicit radiology concepts and CLIP strategy." @default.
- W4386362602 created "2023-09-02" @default.
- W4386362602 creator A5000212230 @default.
- W4386362602 creator A5024796482 @default.
- W4386362602 creator A5026804774 @default.
- W4386362602 creator A5032489574 @default.
- W4386362602 creator A5045468886 @default.
- W4386362602 creator A5046709245 @default.
- W4386362602 creator A5056911306 @default.
- W4386362602 creator A5073053835 @default.
- W4386362602 creator A5083118897 @default.
- W4386362602 date "2023-04-18" @default.
- W4386362602 modified "2023-09-27" @default.
- W4386362602 title "Self-Supervised Learning with Radiology Reports, A Comparative Analysis of Strategies for Large Vessel Occlusion and Brain CTA Images" @default.
- W4386362602 cites W2036200541 @default.
- W4386362602 cites W2194775991 @default.
- W4386362602 cites W2952834869 @default.
- W4386362602 cites W2957996027 @default.
- W4386362602 cites W3031326204 @default.
- W4386362602 cites W3097267201 @default.
- W4386362602 cites W3211074996 @default.
- W4386362602 cites W3214654535 @default.
- W4386362602 cites W4200587271 @default.
- W4386362602 cites W4296027312 @default.
- W4386362602 doi "https://doi.org/10.1109/isbi53787.2023.10230623" @default.
- W4386362602 hasPubMedId "https://pubmed.ncbi.nlm.nih.gov/37711217" @default.
- W4386362602 hasPublicationYear "2023" @default.
- W4386362602 type Work @default.
- W4386362602 citedByCount "0" @default.
- W4386362602 crossrefType "proceedings-article" @default.
- W4386362602 hasAuthorship W4386362602A5000212230 @default.
- W4386362602 hasAuthorship W4386362602A5024796482 @default.
- W4386362602 hasAuthorship W4386362602A5026804774 @default.
- W4386362602 hasAuthorship W4386362602A5032489574 @default.
- W4386362602 hasAuthorship W4386362602A5045468886 @default.
- W4386362602 hasAuthorship W4386362602A5046709245 @default.
- W4386362602 hasAuthorship W4386362602A5056911306 @default.
- W4386362602 hasAuthorship W4386362602A5073053835 @default.
- W4386362602 hasAuthorship W4386362602A5083118897 @default.
- W4386362602 hasConcept C108583219 @default.
- W4386362602 hasConcept C119857082 @default.
- W4386362602 hasConcept C126838900 @default.
- W4386362602 hasConcept C153180895 @default.
- W4386362602 hasConcept C154945302 @default.
- W4386362602 hasConcept C162324750 @default.
- W4386362602 hasConcept C187736073 @default.
- W4386362602 hasConcept C189430467 @default.
- W4386362602 hasConcept C2780451532 @default.
- W4386362602 hasConcept C31601959 @default.
- W4386362602 hasConcept C31972630 @default.
- W4386362602 hasConcept C41008148 @default.
- W4386362602 hasConcept C50644808 @default.
- W4386362602 hasConcept C71924100 @default.
- W4386362602 hasConceptScore W4386362602C108583219 @default.
- W4386362602 hasConceptScore W4386362602C119857082 @default.
- W4386362602 hasConceptScore W4386362602C126838900 @default.
- W4386362602 hasConceptScore W4386362602C153180895 @default.
- W4386362602 hasConceptScore W4386362602C154945302 @default.
- W4386362602 hasConceptScore W4386362602C162324750 @default.
- W4386362602 hasConceptScore W4386362602C187736073 @default.
- W4386362602 hasConceptScore W4386362602C189430467 @default.
- W4386362602 hasConceptScore W4386362602C2780451532 @default.
- W4386362602 hasConceptScore W4386362602C31601959 @default.
- W4386362602 hasConceptScore W4386362602C31972630 @default.
- W4386362602 hasConceptScore W4386362602C41008148 @default.
- W4386362602 hasConceptScore W4386362602C50644808 @default.
- W4386362602 hasConceptScore W4386362602C71924100 @default.
- W4386362602 hasLocation W43863626021 @default.
- W4386362602 hasLocation W43863626022 @default.
- W4386362602 hasOpenAccess W4386362602 @default.
- W4386362602 hasPrimaryLocation W43863626021 @default.
- W4386362602 hasRelatedWork W3014300295 @default.
- W4386362602 hasRelatedWork W3164822677 @default.
- W4386362602 hasRelatedWork W4223943233 @default.
- W4386362602 hasRelatedWork W4225161397 @default.
- W4386362602 hasRelatedWork W4297820521 @default.
- W4386362602 hasRelatedWork W4312200629 @default.
- W4386362602 hasRelatedWork W4360585206 @default.
- W4386362602 hasRelatedWork W4364306694 @default.
- W4386362602 hasRelatedWork W4380075502 @default.
- W4386362602 hasRelatedWork W4380086463 @default.
- W4386362602 isParatext "false" @default.
- W4386362602 isRetracted "false" @default.
- W4386362602 workType "article" @default.