FRINK Linked Data Fragments server

Matches in SemOpenAlex for { <https://semopenalex.org/work/W3205818854> ?p ?o ?g. }

• W3205818854 endingPage "S903" @default.
• W3205818854 startingPage "S902" @default.
• W3205818854 abstract "The use of radiotherapy as part of a multimodality treatment approach has improved outcomes for patients with non-small cell lung cancer (NSCLC), though survival rates continue to lag behind those seen for many other cancers. Survival is frequently assessed using Cox proportional hazards (CPH) regression; however, this method may be too simplistic as it assumes a linear relationship between predictors and the outcome. Deep learning (DL), which employs biologically-inspired artificial neural networks, has been proposed as an alternative method to capture the complex, non-linear associations between variables required for accurate survival prediction. This method is yet to be applied to NSCLC patients undergoing radical radiotherapy, where an accurate predictor of survival would influence patient management. In this retrospective study, we built a DL-based model to predict overall survival using readily available clinical and treatment information on a large dataset of NSCLC patients who received radical radiotherapy and compared its performance against a CPH model. The dataset contained clinical, demographic, treatment and time-to-event survival data for 431 NSCLC patients treated with radical radiotherapy between 2010 and 2015. We built CPH- and DL-based survival prediction models using the following covariates: gender, age, time between diagnosis and treatment, planning target volume, histology, recurrence, chemotherapy and lung volume receiving ≥20 Gy. The DL-based model was implemented using DeepSurv, a deep neural network-based survival prediction algorithm. Data preprocessing included feature scaling of all covariates. The network structure comprised of an input layer, six hidden layers and a binary output layer. The number of nodes in each hidden layer decreased with depth. We used the SeLU activation function, a negative log likelihood CPH-based loss and Nesterov momentum with a learning rate of 2x10-3. We employed L1 and L2 regularisation along with dropout to prevent overfitting. The dataset was randomly divided into mutually exclusive training and testing sets at a ratio of approximately 90:10. We compared the performance of the CPH and DL models using the Harrell's concordance index (c-index) metric. The DL-based survival prediction model generated a c-index of 0.703 after 200 epochs. Training and testing set performance are shown graphically (see figure). The DL method exhibited an improved c-index compared to the conventional CPH method of 0.703 vs 0.637, respectively. We show that, using readily available clinical and treatment variables, DL-based survival analysis demonstrates superior performance over the CPH method for survival prediction in patients with NSCLC undergoing radical radiotherapy." @default.
• W3205818854 created "2021-10-25" @default.
• W3205818854 creator A5000094254 @default.
• W3205818854 creator A5035172314 @default.
• W3205818854 creator A5035844982 @default.
• W3205818854 creator A5056496239 @default.
• W3205818854 creator A5063114847 @default.
• W3205818854 date "2021-10-01" @default.
• W3205818854 modified "2023-10-16" @default.
• W3205818854 title "MA06.03 Deep Learning-Based Survival Prediction for Non-Small Cell Lung Cancer Patients Undergoing Radical Radiotherapy" @default.
• W3205818854 doi "https://doi.org/10.1016/j.jtho.2021.08.136" @default.
• W3205818854 hasPublicationYear "2021" @default.
• W3205818854 type Work @default.
• W3205818854 sameAs 3205818854 @default.
• W3205818854 citedByCount "0" @default.
• W3205818854 crossrefType "journal-article" @default.
• W3205818854 hasAuthorship W3205818854A5000094254 @default.
• W3205818854 hasAuthorship W3205818854A5035172314 @default.
• W3205818854 hasAuthorship W3205818854A5035844982 @default.
• W3205818854 hasAuthorship W3205818854A5056496239 @default.
• W3205818854 hasAuthorship W3205818854A5063114847 @default.
• W3205818854 hasBestOaLocation W32058188541 @default.
• W3205818854 hasConcept C10515644 @default.
• W3205818854 hasConcept C108583219 @default.
• W3205818854 hasConcept C126322002 @default.
• W3205818854 hasConcept C143998085 @default.
• W3205818854 hasConcept C154945302 @default.
• W3205818854 hasConcept C2776256026 @default.
• W3205818854 hasConcept C41008148 @default.
• W3205818854 hasConcept C50382708 @default.
• W3205818854 hasConcept C509974204 @default.
• W3205818854 hasConcept C71924100 @default.
• W3205818854 hasConceptScore W3205818854C10515644 @default.
• W3205818854 hasConceptScore W3205818854C108583219 @default.
• W3205818854 hasConceptScore W3205818854C126322002 @default.
• W3205818854 hasConceptScore W3205818854C143998085 @default.
• W3205818854 hasConceptScore W3205818854C154945302 @default.
• W3205818854 hasConceptScore W3205818854C2776256026 @default.
• W3205818854 hasConceptScore W3205818854C41008148 @default.
• W3205818854 hasConceptScore W3205818854C50382708 @default.
• W3205818854 hasConceptScore W3205818854C509974204 @default.
• W3205818854 hasConceptScore W3205818854C71924100 @default.
• W3205818854 hasIssue "10" @default.
• W3205818854 hasLocation W32058188541 @default.
• W3205818854 hasOpenAccess W3205818854 @default.
• W3205818854 hasPrimaryLocation W32058188541 @default.
• W3205818854 hasRelatedWork W2353103559 @default.
• W3205818854 hasRelatedWork W2357428483 @default.
• W3205818854 hasRelatedWork W2395727201 @default.
• W3205818854 hasRelatedWork W2395953112 @default.
• W3205818854 hasRelatedWork W2568900248 @default.
• W3205818854 hasRelatedWork W3011827526 @default.
• W3205818854 hasRelatedWork W3048057313 @default.
• W3205818854 hasRelatedWork W3189109524 @default.
• W3205818854 hasRelatedWork W3197588893 @default.
• W3205818854 hasRelatedWork W1912371934 @default.
• W3205818854 hasVolume "16" @default.
• W3205818854 isParatext "false" @default.
• W3205818854 isRetracted "false" @default.
• W3205818854 magId "3205818854" @default.
• W3205818854 workType "article" @default.