ISeeU: Visually interpretable deep learning for mortality prediction inside the ICU

To improve the performance of Intensive Care Units (ICUs), the field of bio-statistics has developed scores which try to predict the likelihood of negative outcomes. These help evaluate the effectiveness of treatments and clinical practice, and also help to identify patients with unexpected outcomes...

Full description

Autores:

Tipo de recurso:

Fecha de publicación:: 2019

Institución:: Universidad Tecnológica de Bolívar

Repositorio:: Repositorio Institucional UTB

Idioma:: eng

id	UTB2_5dd5a420076ab3c2709b41b282a25926
oai_identifier_str	oai:repositorio.utb.edu.co:20.500.12585/8995
network_acronym_str	UTB2
network_name_str	Repositorio Institucional UTB
repository_id_str
dc.title.none.fl_str_mv	ISeeU: Visually interpretable deep learning for mortality prediction inside the ICU
title	ISeeU: Visually interpretable deep learning for mortality prediction inside the ICU
spellingShingle	ISeeU: Visually interpretable deep learning for mortality prediction inside the ICU Deep learning ICU MIMIC-III Shapley values Deep learning Forecasting Game theory Intensive care units Clinical practices Coalitional game theory Medical information MIMIC-III Relevant features Shapley value Sub-optimal performance Traditional techniques Deep neural networks Adoption Article Deep learning Game Human Intensive care unit Medical information Mortality Prediction Deep neural network
title_short	ISeeU: Visually interpretable deep learning for mortality prediction inside the ICU
title_full	ISeeU: Visually interpretable deep learning for mortality prediction inside the ICU
title_fullStr	ISeeU: Visually interpretable deep learning for mortality prediction inside the ICU
title_full_unstemmed	ISeeU: Visually interpretable deep learning for mortality prediction inside the ICU
title_sort	ISeeU: Visually interpretable deep learning for mortality prediction inside the ICU
dc.subject.keywords.none.fl_str_mv	Deep learning ICU MIMIC-III Shapley values Deep learning Forecasting Game theory Intensive care units Clinical practices Coalitional game theory Medical information MIMIC-III Relevant features Shapley value Sub-optimal performance Traditional techniques Deep neural networks Adoption Article Deep learning Game Human Intensive care unit Medical information Mortality Prediction Deep neural network
topic	Deep learning ICU MIMIC-III Shapley values Deep learning Forecasting Game theory Intensive care units Clinical practices Coalitional game theory Medical information MIMIC-III Relevant features Shapley value Sub-optimal performance Traditional techniques Deep neural networks Adoption Article Deep learning Game Human Intensive care unit Medical information Mortality Prediction Deep neural network
description	To improve the performance of Intensive Care Units (ICUs), the field of bio-statistics has developed scores which try to predict the likelihood of negative outcomes. These help evaluate the effectiveness of treatments and clinical practice, and also help to identify patients with unexpected outcomes. However, they have been shown by several studies to offer sub-optimal performance. Alternatively, Deep Learning offers state of the art capabilities in certain prediction tasks and research suggests deep neural networks are able to outperform traditional techniques. Nevertheless, a main impediment for the adoption of Deep Learning in healthcare is its reduced interpretability, for in this field it is crucial to gain insight into the why of predictions, to assure that models are actually learning relevant features instead of spurious correlations. To address this, we propose a deep multi-scale convolutional architecture trained on the Medical Information Mart for Intensive Care III (MIMIC-III) for mortality prediction, and the use of concepts from coalitional game theory to construct visual explanations aimed to show how important these inputs are deemed by the network. Results show our model attains a ROC AUC of 0.8735 (± 0.0025) which is competitive with the state of the art of Deep Learning mortality models trained on MIMIC-III data, while remaining interpretable. Supporting code can be found at https://github.com/williamcaicedo/ISeeU. © 2019 Elsevier Inc.
publishDate	2019
dc.date.issued.none.fl_str_mv	2019
dc.date.accessioned.none.fl_str_mv	2020-03-26T16:32:44Z
dc.date.available.none.fl_str_mv	2020-03-26T16:32:44Z
dc.type.coarversion.fl_str_mv	http://purl.org/coar/version/c_970fb48d4fbd8a85
dc.type.coar.fl_str_mv	http://purl.org/coar/resource_type/c_2df8fbb1
dc.type.driver.none.fl_str_mv	info:eu-repo/semantics/article
dc.type.hasVersion.none.fl_str_mv	info:eu-repo/semantics/publishedVersion
dc.type.spa.none.fl_str_mv	Artículo
status_str	publishedVersion
dc.identifier.citation.none.fl_str_mv	Journal of Biomedical Informatics; Vol. 98
dc.identifier.issn.none.fl_str_mv	15320464
dc.identifier.uri.none.fl_str_mv	https://hdl.handle.net/20.500.12585/8995
dc.identifier.doi.none.fl_str_mv	10.1016/j.jbi.2019.103269
dc.identifier.instname.none.fl_str_mv	Universidad Tecnológica de Bolívar
dc.identifier.reponame.none.fl_str_mv	Repositorio UTB
dc.identifier.orcid.none.fl_str_mv	55782426500 57211703831
identifier_str_mv	Journal of Biomedical Informatics; Vol. 98 15320464 10.1016/j.jbi.2019.103269 Universidad Tecnológica de Bolívar Repositorio UTB 55782426500 57211703831
url	https://hdl.handle.net/20.500.12585/8995
dc.language.iso.none.fl_str_mv	eng
language	eng
dc.rights.coar.fl_str_mv	http://purl.org/coar/access_right/c_16ec
dc.rights.uri.none.fl_str_mv	http://creativecommons.org/licenses/by-nc-nd/4.0/
dc.rights.accessRights.none.fl_str_mv	info:eu-repo/semantics/restrictedAccess
dc.rights.cc.none.fl_str_mv	Atribución-NoComercial 4.0 Internacional
rights_invalid_str_mv	http://creativecommons.org/licenses/by-nc-nd/4.0/ Atribución-NoComercial 4.0 Internacional http://purl.org/coar/access_right/c_16ec
eu_rights_str_mv	restrictedAccess
dc.format.medium.none.fl_str_mv	Recurso electrónico
dc.format.mimetype.none.fl_str_mv	application/pdf
dc.publisher.none.fl_str_mv	Academic Press Inc.
publisher.none.fl_str_mv	Academic Press Inc.
dc.source.none.fl_str_mv	https://www.scopus.com/inward/record.uri?eid=2-s2.0-85070867164&doi=10.1016%2fj.jbi.2019.103269&partnerID=40&md5=b824faf402f646458dd0679ca76fb069
institution	Universidad Tecnológica de Bolívar
bitstream.url.fl_str_mv	https://repositorio.utb.edu.co/bitstream/20.500.12585/8995/1/MiniProdInv.png
bitstream.checksum.fl_str_mv	0cb0f101a8d16897fb46fc914d3d7043
bitstream.checksumAlgorithm.fl_str_mv	MD5
repository.name.fl_str_mv	Repositorio Institucional UTB
repository.mail.fl_str_mv	repositorioutb@utb.edu.co
_version_	1814021654303473664
spelling	2020-03-26T16:32:44Z2020-03-26T16:32:44Z2019Journal of Biomedical Informatics; Vol. 9815320464https://hdl.handle.net/20.500.12585/899510.1016/j.jbi.2019.103269Universidad Tecnológica de BolívarRepositorio UTB5578242650057211703831To improve the performance of Intensive Care Units (ICUs), the field of bio-statistics has developed scores which try to predict the likelihood of negative outcomes. These help evaluate the effectiveness of treatments and clinical practice, and also help to identify patients with unexpected outcomes. However, they have been shown by several studies to offer sub-optimal performance. Alternatively, Deep Learning offers state of the art capabilities in certain prediction tasks and research suggests deep neural networks are able to outperform traditional techniques. Nevertheless, a main impediment for the adoption of Deep Learning in healthcare is its reduced interpretability, for in this field it is crucial to gain insight into the why of predictions, to assure that models are actually learning relevant features instead of spurious correlations. To address this, we propose a deep multi-scale convolutional architecture trained on the Medical Information Mart for Intensive Care III (MIMIC-III) for mortality prediction, and the use of concepts from coalitional game theory to construct visual explanations aimed to show how important these inputs are deemed by the network. Results show our model attains a ROC AUC of 0.8735 (± 0.0025) which is competitive with the state of the art of Deep Learning mortality models trained on MIMIC-III data, while remaining interpretable. Supporting code can be found at https://github.com/williamcaicedo/ISeeU. © 2019 Elsevier Inc.Recurso electrónicoapplication/pdfengAcademic Press Inc.http://creativecommons.org/licenses/by-nc-nd/4.0/info:eu-repo/semantics/restrictedAccessAtribución-NoComercial 4.0 Internacionalhttp://purl.org/coar/access_right/c_16echttps://www.scopus.com/inward/record.uri?eid=2-s2.0-85070867164&doi=10.1016%2fj.jbi.2019.103269&partnerID=40&md5=b824faf402f646458dd0679ca76fb069ISeeU: Visually interpretable deep learning for mortality prediction inside the ICUinfo:eu-repo/semantics/articleinfo:eu-repo/semantics/publishedVersionArtículohttp://purl.org/coar/version/c_970fb48d4fbd8a85http://purl.org/coar/resource_type/c_2df8fbb1Deep learningICUMIMIC-IIIShapley valuesDeep learningForecastingGame theoryIntensive care unitsClinical practicesCoalitional game theoryMedical informationMIMIC-IIIRelevant featuresShapley valueSub-optimal performanceTraditional techniquesDeep neural networksAdoptionArticleDeep learningGameHumanIntensive care unitMedical informationMortalityPredictionDeep neural networkCaicedo-Torres W.Gutierrez J.Johnson, A.E.W., Ghassemi, M.M., Nemati, S., Niehaus, K.E., Clifton, D., Clifford, G.D., Machine learning and decision support in critical care (2016) Proc. IEEE, 104 (2), pp. 444-466Rapsang, A.G., Shyam, D.C., Scoring systems in the intensive care unit: a compendium (2014) Indian J. Crit. Care Med.: Peer-Review., Off. Publicat. Indian Soc. Crit. Care Med., 18 (4), pp. 220-228. , http://www.ncbi.nlm.nih.gov/pmc/articles/PMC4033855/Huang, Y.C., Chang, K.Y., Lin, S.P., Chen, K., Chan, K.H., Chang, P., Development of a daily mortality probability prediction model from Intensive Care Unit patients using a discrete-time event history analysis (2013) Comput. Methods Progr. Biomed., 111 (2), pp. 280-289. , http://www.sciencedirect.com/science/article/pii/S0169260713001107Paul, E., Bailey, M., Pilcher, D., Risk prediction of hospital mortality for adult patients admitted to Australian and New Zealand intensive care units: development and validation of the Australian and New Zealand Risk of Death model (2013) J. Crit. Care, 28 (6), pp. 935-941. , http://www.sciencedirect.com/science/article/pii/S0883944113002566Che, Z., Purushotham, S., Khemani, R., Liu, Y., Interpretable deep models for ICU outcome prediction (2016) AMIA Annu. Symp. Proc., 2016, pp. 371-380. , http://www.ncbi.nlm.nih.gov/pmc/articles/PMC5333206/Lipton, Z.C., (2016), http://arxiv.org/abs/1606.03490, The Mythos of Model Interpretability, ICML Workshop on Human Interpretability in Machine Learning abs/1606.0 96–100. arXiv:arXiv:1606.03490v1He, K., Zhang, X., Ren, S., Sun, J., Deep residual learning for image (2016) Recognition, 2016 IEEE Conference on Computer Vision and Pattern Recognition (CVPR), pp. 770-778Szegedy, C., Liu, W., Jia, Y., Sermanet, P., Reed, S., Anguelov, D., Erhan, D., Rabinovich, A., Going deeper with convolutions (2015), pp. 1-9. , 2015 IEEE Conference on Computer Vision and Pattern Recognition (CVPR)Urban, G., Geras, K.J., Kahou, S.E., Aslan, O., Wang, S., Caruana, R., (2017), A.-r. Mohamed, M. Philipose, M. Richardson, Do Deep Convolutional Nets Really Need to be Deep (Or Even Convolutional)?, in: ICLRCooper, G.F., Aliferis, C.F., Ambrosino, R., Aronis, J., Buchanan, B.G., Caruana, R., Fine, M.J., Spirtes, P., An evaluation of machine-learning methods for predicting pneumonia mortality, Artif. Intell. MedShapley, L.S., A Value for n-Person Games (1953) Contributions to the Theory of Games II, pp. 307-317. , H.W. Kuhn A.W. Tucker Princeton University Press PrincetonChe, Z., Purushotham, S., Cho, K., Sontag, D., Liu, Y., Recurrent Neural Networks for Multivariate Time Series with Missing Values, CoRR abs/1606.0Grnarova, P., Schmidt, F., Hyland, S.L., Eickhoff, C., Neural Document Embeddings for Intensive Care Patient Mortality Prediction, CoRR abs/1612.0Purushotham, S., Meng, C., Che, Z., Liu, Y., http://www.sciencedirect.com/science/article/pii/S1532046418300716, Benchmarking deep learning models on large healthcare datasets, J. Biomed. Inform. https://doi.org/10.1016/j.jbi.2018.04.007Ba, L.J., Caruana, R., Do Deep Nets Really Need to Be Deep? (2014) Proceedings of the 27th International Conference on Neural Information Processing Systems, NIPS'14, pp. 2654-2662. , http://dl.acm.org/citation.cfm?id=2969033.2969123, MIT Press Cambridge, MA, USASuresh, H., Hunt, N., Johnson, A., Celi, L.A., Szolovits, P., Ghassemi, M., Clinical Intervention Prediction and Understanding using Deep Networks, arXiv preprint arXiv:1705.08498Shen, D., Wu, G., Suk, H.-I., Deep learning in medical image analysis (2017) Annu. Rev. Biomed. Eng., 19 (1), p. nullLasko, T.A., Denny, J.C., Levy, M.A., Computational phenotype discovery using unsupervised feature learning over noisy, sparse, and irregular clinical data (2013) PLoS ONE, 8 (6), p. e66341. , http://www.ncbi.nlm.nih.gov/pmc/articles/PMC3691199/Che, Z., Kale, D., Li, W., Bahadori, M.T., Liu, Y., Deep Computational Phenotyping (2015), pp. 507-516. , http://doi.acm.org/10.1145/2783258.2783365, https://doi.org/10.1145/2783258.2783365. Proceedings of the 21th ACM SIGKDD International Conference on Knowledge Discovery and Data Mining, KDD ’15, ACM, New York, NY, USASilva, I., Moody, G., Scott, D.J., Celi, L.A., Mark, R.G., Predicting In-Hospital Mortality of ICU Patients: the PhysioNet/Computing in Cardiology Challenge 2012 (2012) Comput. Cardiol., 39, pp. 245-248. , http://www.ncbi.nlm.nih.gov/pmc/articles/PMC3965265/Kale, D.C., Che, Z., Bahadori, M.T., Li, W., Liu, Y., Wetzel, R., Causal phenotype discovery via deep networks (2015) AMIA Annu. Symp. Proc., 2015, pp. 677-686. , http://www.ncbi.nlm.nih.gov/pmc/articles/PMC4765623/Lipton, Z.C., Kale, D.C., Elkan, C., Wetzell, R., Learning to Diagnose with LSTM Recurrent Neural Networks (2015) International Conference on Learning Representations (ICLR 2016)Hochreiter, S., Schmidhuber, J., Long short-term memory (1997) Neural Comput., 9 (8), pp. 1735-1780Lipton, Z.C., Kale, D., Wetzel, R., Doshi-Velez, F., Fackler, J., Kale, D., Wallace, B., Directly Modeling Missing Data in Sequences with RNNs: Improved Classification of Clinical Time Series (2016), 56, pp. 253-270. , http://proceedings.mlr.press/v56/Lipton16.html, of Proceedings of Machine Learning Research, PMLR, Northeastern University, Boston, MA, USA J. Weins (Eds.), Proceedings of the 1st Machine Learning for Healthcare ConferenceRazavian, N., Sontag, D., (2015), Temporal convolutional neural networks for diagnosis lab tests, 25 November abs/1511.0 1–17. arXiv:1151.07938v1Johnson, A.E.W., Pollard, T.J., Shen, L., Lehman, L.-W.H., Feng, M., Ghassemi, M., Moody, B., Mark, R.G., MIMIC-III, a freely accessible critical care database (2016) Sci. Data, 3, p. 160035Rumelhart, D.E., Hinton, G.E., Williams, R.J., Learning representations by back-propagating errors (1986) Nature, 323 (6088), pp. 533-536Choi, E., Bahadori, M.T., Song, L., Stewart, W.F., Sun, J., GRAM: Graph-based Attention Model for Healthcare Representation Learning, CoRR abs/1611.0Aczon, M., Ledbetter, D., Ho, L., Gunny, A., Flynn, A., Williams, J., Wetzel, R., Dynamic mortality risk predictions in pediatric critical care using recurrent (2017) Neural Networks, , arXiv:1701.06675Beaulieu-Jones, B.K., Orzechowski, P., Moore, J.H., Mapping patient trajectories using longitudinal extraction and deep learning in the MIMIC-III critical care database (2018) Pacific Symposium on Biocomputing. Pacific Symposium on Biocomputing, pp. 123-132. , 23Gall, J.R., Lemeshow, S., Saulnier, F., A New Simplified Acute Physiology Score (SAPS II) Based on a European/North American Multicenter Study, JAMA: J. Am. Med. Assoc. arXiv:0402594v3Calvert, J., Mao, Q., Hoffman, J.L., Jay, M., Desautels, T., Mohamadlou, H., Chettipally, U., Das, R., Using electronic health record collected clinical variables to predict medical intensive care unit mortality, Annals Med. SurgLeCun, Y., Bottou, L., Bengio, Y., Haffner, Gradient-Based Learning Applied to Document Recognition (1998), 86, pp. 2278-2324. , Proceedings of the IEEERumelhart, D.E., Hinton, G.E., Williams, R.J., Learning representations by back-propagating errors, NaturearXiv:arXiv:1011.1669v3Goodfellow, I., Bengio, Y., Courville, A., Deep Learning (2016), MIT PressShapiro, L., Computer Vision (2001), Prentice Hall Upper Saddle River, NJSrivastava, N., Hinton, G., Krizhevsky, A., Sutskever, I., Salakhutdinov, R., Dropout: a simple way to prevent neural networks from overfitting (2014) J. Mach. Learn. Res., 15 (1), pp. 1929-1958. , http://dl.acm.org/citation.cfm?id=2627435.2670313Ioffe, S., Szegedy, C., (2015), Batch Normalization: Accelerating Deep Network Training by Reducing Internal Covariate Shift, Arxiv abs/1502.0 1–11. arXiv:1502.03167Strumbelj, E., Kononenko, I., Wrobel, S., An efficient explanation of individual classifications using game theory, J. Mach. Learn. Res. arXiv:1606.05386Zeiler, M.D., Fergus, R., Visualizing and Understanding Convolutional Networks (2014), pp. 818-833. , Springer International Publishing ChamShrikumar, A., Greenside, P., Kundaje, A., Learning Important Features Through Propagating Activation Differences, CoRR abs/1704.0. arXiv:1704.02685Bach, S., Binder, A., Montavon, G., Klauschen, F., Müller, K.R., Samek, W., On pixel-wise explanations for non-linear classifier decisions by layer-wise relevance propagation, PLoS ONEarXiv:1606.04155Simonyan, K., Vedaldi, A., Zisserman, A., Deep Inside Convolutional Networks: Visualising Image Classification Models and Saliency Maps, CoRR abs/1312.6Springenberg, J.T., Dosovitskiy, A., Brox, T., Riedmiller, M.A., Striving for Simplicity: The All Convolutional Net, CoRR abs/1412.6. arXiv:1412.6806Kindermans, P.-J., Schütt, K., Müller, K.-R., Dähne, S., Investigating the influence of noise and distractors on the interpretation of neural networks, ArXiv e-prints arXiv:1611.07270Lundberg, S.M., Lee, S.-I., Guyon, I., Luxburg, U.V., Bengio, S., Wallach, H., Fergus, R., Vishwanathan, S., A Unified Approach to Interpreting Model Predictions (2017), pp. 4765-4774. , http://papers.nips.cc/paper/7062-a-unified-approach-to-interpreting-model-predictions.pdf, R. Garnett (Eds.), Advances in Neural Information Processing Systems 30, Curran Associates IncChollet, F., Keras 2015Abadi, M., Agarwal, A., Barham, P., Brevdo, E., Chen, Z., Citro, C., Corrado, G., Zheng, X., (2015), http://download.tensorflow.org/paper/whitepaper2015.pdf, TensorFlow: Large-Scale Machine Learning on Heterogeneous Distributed Systems, None 1 (212) 19. arXiv:1603.04467, https://doi.org/10.1038/nn.3331Johnson, A.E., Stone, D.J., Celi, L.A., Pollard, T.J., The, M.I.M.I., C Code Repository: Enabling reproducibility in critical care research, J. Am. Med. Inform. Assochttp://purl.org/coar/resource_type/c_6501THUMBNAILMiniProdInv.pngMiniProdInv.pngimage/png23941https://repositorio.utb.edu.co/bitstream/20.500.12585/8995/1/MiniProdInv.png0cb0f101a8d16897fb46fc914d3d7043MD5120.500.12585/8995oai:repositorio.utb.edu.co:20.500.12585/89952021-02-02 15:10:18.927Repositorio Institucional UTBrepositorioutb@utb.edu.co

ISeeU: Visually interpretable deep learning for mortality prediction inside the ICU

Publicaciones similares