Cargando…
2438. Using Artificial Neural Networks to Predict Intra-Abdominal Abscess Risk Post-Appendectomy
BACKGROUND: Applying Artificial Intelligence techniques to healthcare data are gaining momentum. Early identification of patients at risk of surgical site infections is a major clinical goal. Our objective for this study was to determine whether deep learning AI techniques could identify patients at...
Autores principales: | , , , , |
---|---|
Formato: | Online Artículo Texto |
Lenguaje: | English |
Publicado: |
Oxford University Press
2019
|
Materias: | |
Acceso en línea: | https://www.ncbi.nlm.nih.gov/pmc/articles/PMC6810183/ http://dx.doi.org/10.1093/ofid/ofz360.2116 |
_version_ | 1783462187427168256 |
---|---|
author | Al Khatib, Hassan S Alramadhan, Morouge Murphy, James Tsao, KuoJen Chang, Michael L |
author_facet | Al Khatib, Hassan S Alramadhan, Morouge Murphy, James Tsao, KuoJen Chang, Michael L |
author_sort | Al Khatib, Hassan S |
collection | PubMed |
description | BACKGROUND: Applying Artificial Intelligence techniques to healthcare data are gaining momentum. Early identification of patients at risk of surgical site infections is a major clinical goal. Our objective for this study was to determine whether deep learning AI techniques could identify patients at risk of intra-abdominal abscess development post-appendectomy using clinical data for pediatric patients undergoing appendectomy. METHODS: A dataset of 1,574 patients classified by surgeons as negative (1,328) or positive (246) for Intra-Abdominal Abscess Post-Appendectomy for Appendicitis were selected from a database containing 6,127 patients less than 19 years-old who had appendectomy at our institution between 2009–2018. Demographic, clinical, and surgical information were extracted. 34 Independent variables were identified to be useful for the study. Using Random Forest methodology 12 variables with the highest influence on the outcome were selected for the final dataset. Data imputation (MICE algorithm) was used to replace missing data points. Two “Reproducible” Artificial Neural Networks with different architectures were developed to predict the risk of developing Intra-Abdominal Abscess Post- Appendectomy: Model (1) 12 Inputs, 3 hidden layers with 12 Neurons each, and 1 Output. Model (2) 12 Inputs, 2 hidden layers with 18 Neurons each, and 1 Output. RESULTS: For the 1,574 patients (80%-20% used as training and test sets), Model (1) achieved Accuracy of 89.84%, Sensitivity of ~ 70%, and Specificity of 93.61% on the test set while Model (2) achieved Accuracy of 84.13%, Sensitivity of 81.63%, and Specificity of 84.6%. The difference between the models is that in Model (2) we over sampled the minority class (using SMOTE algorithm) to balance both classes which helped the model to learn both classes without bias and improved sensitivity over Model (1). CONCLUSION: Deep learning algorithms applied to enough clinical variables can identify patients with high probability for the risk of developing intra-abdominal abscess post-appendectomy. While further test sets are necessary to validate the models, Artificial Neural Networks can be an important addition to current post-surgical care guidelines to personalize and optimize care to reduce infections following appendectomy. [Image: see text] [Image: see text] [Image: see text] DISCLOSURES: All authors: No reported disclosures. |
format | Online Article Text |
id | pubmed-6810183 |
institution | National Center for Biotechnology Information |
language | English |
publishDate | 2019 |
publisher | Oxford University Press |
record_format | MEDLINE/PubMed |
spelling | pubmed-68101832019-10-28 2438. Using Artificial Neural Networks to Predict Intra-Abdominal Abscess Risk Post-Appendectomy Al Khatib, Hassan S Alramadhan, Morouge Murphy, James Tsao, KuoJen Chang, Michael L Open Forum Infect Dis Abstracts BACKGROUND: Applying Artificial Intelligence techniques to healthcare data are gaining momentum. Early identification of patients at risk of surgical site infections is a major clinical goal. Our objective for this study was to determine whether deep learning AI techniques could identify patients at risk of intra-abdominal abscess development post-appendectomy using clinical data for pediatric patients undergoing appendectomy. METHODS: A dataset of 1,574 patients classified by surgeons as negative (1,328) or positive (246) for Intra-Abdominal Abscess Post-Appendectomy for Appendicitis were selected from a database containing 6,127 patients less than 19 years-old who had appendectomy at our institution between 2009–2018. Demographic, clinical, and surgical information were extracted. 34 Independent variables were identified to be useful for the study. Using Random Forest methodology 12 variables with the highest influence on the outcome were selected for the final dataset. Data imputation (MICE algorithm) was used to replace missing data points. Two “Reproducible” Artificial Neural Networks with different architectures were developed to predict the risk of developing Intra-Abdominal Abscess Post- Appendectomy: Model (1) 12 Inputs, 3 hidden layers with 12 Neurons each, and 1 Output. Model (2) 12 Inputs, 2 hidden layers with 18 Neurons each, and 1 Output. RESULTS: For the 1,574 patients (80%-20% used as training and test sets), Model (1) achieved Accuracy of 89.84%, Sensitivity of ~ 70%, and Specificity of 93.61% on the test set while Model (2) achieved Accuracy of 84.13%, Sensitivity of 81.63%, and Specificity of 84.6%. The difference between the models is that in Model (2) we over sampled the minority class (using SMOTE algorithm) to balance both classes which helped the model to learn both classes without bias and improved sensitivity over Model (1). CONCLUSION: Deep learning algorithms applied to enough clinical variables can identify patients with high probability for the risk of developing intra-abdominal abscess post-appendectomy. While further test sets are necessary to validate the models, Artificial Neural Networks can be an important addition to current post-surgical care guidelines to personalize and optimize care to reduce infections following appendectomy. [Image: see text] [Image: see text] [Image: see text] DISCLOSURES: All authors: No reported disclosures. Oxford University Press 2019-10-23 /pmc/articles/PMC6810183/ http://dx.doi.org/10.1093/ofid/ofz360.2116 Text en © The Author(s) 2019. Published by Oxford University Press on behalf of Infectious Diseases Society of America. http://creativecommons.org/licenses/by-nc-nd/4.0/ This is an Open Access article distributed under the terms of the Creative Commons Attribution-NonCommercial-NoDerivs licence (http://creativecommons.org/licenses/by-nc-nd/4.0/), which permits non-commercial reproduction and distribution of the work, in any medium, provided the original work is not altered or transformed in any way, and that the work is properly cited. For commercial re-use, please contact journals.permissions@oup.com |
spellingShingle | Abstracts Al Khatib, Hassan S Alramadhan, Morouge Murphy, James Tsao, KuoJen Chang, Michael L 2438. Using Artificial Neural Networks to Predict Intra-Abdominal Abscess Risk Post-Appendectomy |
title | 2438. Using Artificial Neural Networks to Predict Intra-Abdominal Abscess Risk Post-Appendectomy |
title_full | 2438. Using Artificial Neural Networks to Predict Intra-Abdominal Abscess Risk Post-Appendectomy |
title_fullStr | 2438. Using Artificial Neural Networks to Predict Intra-Abdominal Abscess Risk Post-Appendectomy |
title_full_unstemmed | 2438. Using Artificial Neural Networks to Predict Intra-Abdominal Abscess Risk Post-Appendectomy |
title_short | 2438. Using Artificial Neural Networks to Predict Intra-Abdominal Abscess Risk Post-Appendectomy |
title_sort | 2438. using artificial neural networks to predict intra-abdominal abscess risk post-appendectomy |
topic | Abstracts |
url | https://www.ncbi.nlm.nih.gov/pmc/articles/PMC6810183/ http://dx.doi.org/10.1093/ofid/ofz360.2116 |
work_keys_str_mv | AT alkhatibhassans 2438usingartificialneuralnetworkstopredictintraabdominalabscessriskpostappendectomy AT alramadhanmorouge 2438usingartificialneuralnetworkstopredictintraabdominalabscessriskpostappendectomy AT murphyjames 2438usingartificialneuralnetworkstopredictintraabdominalabscessriskpostappendectomy AT tsaokuojen 2438usingartificialneuralnetworkstopredictintraabdominalabscessriskpostappendectomy AT changmichaell 2438usingartificialneuralnetworkstopredictintraabdominalabscessriskpostappendectomy |