Comparison of Brain Injury Mortality Prediction by Machine Learning Models: logistic regression and support vector machine. Systematic review and meta-analysis

Comparação da Previsão de Mortalidade por Trauma Cranioencefálico pelos Modelos de Aprendizado de Máquina: regressão logística e máquina de vetores de suporte. Revisão sistemática e meta-análise

• Samanttha Cristina da Silva Chaves (1)    
• Isadhora Maria Maran de Souza (2)
• Julia Augusta Guimarães Dourado (3)    
• Henrique Lico de Souza (4)
• Eduardo José Domingues (5)
• Adilson José Manuel de Oliveira (6)

• https://doi.org/10.22290/jbnc.2023.340112
• J Bras Neurocirur 34 (1): 107-112, 2023

Resumo

Introdução: Com o advento da inteligência artificial (IA) surgiram vários modelos de aprendizado de máquina médico. Como o tempo para atendimento médico adequado é fundamental, este artigo busca confrontar dados e apontar qual dos modelos de aprendizado de máquina – Regressão Logística (LR) ou Máquina de Vetores de Suporte (SVM) – para lesão cerebral é mais assertivo e oferece maior segurança ao médico para tomar decisões terapêuticas rápidas. Objetivos: Avaliar o melhor modelo de aprendizado de máquina para prever mortalidade por TCE. Método: A busca foi realizada nas bases de dados PubMed, BVS, Lilacs e Scielo. Três artigos comparativos dos modelos de aprendizado de máquina Regressão Logística e Máquina de Vetor de Suporte foram submetidos à revisão sistemática e meta-análise, e quatro valores de parâmetros de análise que predizem mortalidade por traumatismo cranioencefálico foram comparados. Resultado: Para a “Área sob a curva” (AUC) e “Precisão”, o modelo de Regressão Logística (RL) superou o modelo de Máquina de Vetor de Suporte (MVS). Para os parâmetros “sensibilidade” e “precisão”, denota-se que a MVS superou a RL. Conclusão: Mais estudos, e maiores, devem ser realizados a fim de quantificar qual modelo de aprendizado de máquina é mais sensível para predizer mortalidade por TCE.

Palavras-chave

Traumatismo cranioencefálico; Aprendizado de Máquina; Máquina de vetor de suporte; Regressão logística; Predição de mortalidade

Abstract

Introduction: Following the advent of artificial intelligence (AI), a number of medical machine learning models have emerged. Since timing for adequate medical care is fundamental, this article seeks to confront data and point out which of the machine learning models – Logistic Regression (LR) or Support Vector Machine (SVM) – for brain injury is more assertive and provides greater assurance to the physician to make quick therapeutic decisions. Objectives: to evaluate the best machine learning model to predict TBI mortality. Methods: The search was conducted in PubMed, BVS, Lilacs and Scielo databases. Three comparative articles of the machine learning models Logistic Regression and Support Vector Machine were submitted to the systematic review and meta-analysis, and four values of analysis parameters predicting mortality from head trauma were compared. Result: For ‘Area under the curve’ (AUC) and ‘Accuracy’, the Logistic Regression (LR) model outperformed the Support Vector Machine (SVM) model. For ‘sensitivity’ and ‘precision’ parameters, it is denoted that the SVM outperformed the LR. Conclusion: More and larger studies should be carried out in order to quantify which machine learning model is most sensitive to predict mortality from TBI.

Keywords

Traumatic brain injury; Machine learning; Support vector machine; Logistic regression; Mortality prediction

References

1. Raj R, Luostarinen T, Pursiainen E, et al. Machine learning-based dynamic mortality prediction after traumatic brain injury. Sci Rep. 2019;9(1):17672. http://dx.doi.org/10.1038/s41598-019-53889-6. PMid:31776366.

2. Matsuo K, Aihara H, Nakai T, Morishita A, Tohma Y, Kohmura E. Machine learning to predict in-hospital morbidity and mortality after traumatic brain injury. J Neurotrauma. 2020;37(1):202-10. http://dx.doi. org/10.1089/neu.2018.6276. PMid:31359814.

3. Ahmed FS, Ali L, Joseph BA, Ikram A, Ul Mustafa R, Bukhari SAC. A statistically rigorous deep neural network approach to predict mortality in trauma patients admitted to the intensive care unit. J Trauma Acute Care Surg. 2020;89(4):736-42. http://dx.doi.org/10.1097/ TA.0000000000002888. PMid:32773672.

4. Feng J, Wang Y, Peng J, Sun M, Zeng J, Jiang H. Comparison between logistic regression and machine learning algorithms on survival prediction of traumatic brain injuries. J Crit Care. 2019;54:110-6. http:// dx.doi.org/10.1016/j.jcrc.2019.08.010. PMid:31408805.

5. Rau C, Kuo P, Chien P, Huang C, Hsieh H, Hsieh C. Mortality prediction in patients with isolated moderate and severe traumatic brain injury using machine learning models. PLoS One. 2018;13(11):e0207192. http://dx.doi.org/10.1371/journal.pone.0207192. PMid:30412613.

6. Abujaber A, Fadlalla A, Gammoh D, Abdelrahman H, Mollazehi M, El-Menyar A. Prediction of in-hospital mortality in patients with post traumatic brain injury using National Trauma Registry and Machine Learning Approach. Scand J Trauma Resusc Emerg Med. 2020;28(1):28- 44. http://dx.doi.org/10.1186/s13049-020-00738-5. PMid:32460867.

7. Zreik J, Goyal A, Bhandarkar AR, et al. The utility of machine learning algorithms in predicting in-hospital mortality following traumatic brain injuries. Insights from the national trauma databank. Neurosurgery. 2020(Suppl 1):141. http://dx.doi.org/10.1093/neuros/nyaa447_420.

8. Ji SY, Smith R, Huynh T, Najarian K. A comparative analysis of multi-level computer-assisted decision making systems for traumatic injuries. BMC Med Inform Decis Mak. 2009;9:2. http://dx.doi.org/10.1186/1472- 6947-9-2. PMid:19144188.

9. Gravesteijn BY, Nieboer D, Ercole A, et al. Machine learning algorithms performed no better than regression models for prognostication in traumatic brain injury. J Clin Epidemiol. 2020;122:95- 107. http://dx.doi.org/10.1016/j.jclinepi.2020.03.005. PMid:32201256.

10. Serviá L, Montserrat N, Badia M, et al. Machine learning techniques for mortality prediction in critical traumatic patients: anatomic and physiologic variables from the RETRAUCI study. BMC Med Res Methodol. 2020;20(1):262. http://dx.doi.org/10.1186/s12874-020- 01151-3. PMid:33081694.

11. Kayhanian S, Young A, Mangla C, et al. Modelling outcomes after paediatric brain injury with admission laboratory values: a machine-learning approach. Pediatr Res. 2019;86(5):641-5. http://dx.doi. org/10.1038/s41390-019-0510-9. PMid:31349360.

12. van der Ploeg T, Nieboer D, Steyerberg EW. Modern modeling techniques had limited external validity in predicting mortality from traumatic brain injury. J Clin Epidemiol. 2016;78:83-9. http://dx.doi. org/10.1016/j.jclinepi.2016.03.002. PMid:26987507.

13. Wu E, Marthi S, Asaad W. Predictors of mortality in traumatic intracranial hemorrhage: a National Trauma Data Bank study. Front Neurol. 2020;11:587587. http://dx.doi.org/10.3389/fneur.2020.587587. PMid:33281725.

14. Sefrioui I, Amadini R, Mauro J, El Fallahi A, Gabbrielli M. Survival prediction of trauma patients: a study on US National Trauma Data Bank. Eur J Trauma Emerg Surg. 2017;43(6):805-22. http://dx.doi. org/10.1007/s00068-016-0757-3. PMid:28229175.

15. Gorczyca M, Toscano N, Cheng J. The trauma severity model: an ensemble machine learning approach to risk prediction. Comp Biol Med. 2019:108:9-19. https://doi.org/10.1016/j.compbiomed.2019.02.025.

---

1 Medicine student, Biotechnology Institute, Federal University of Catalão – UFCAT, Catalão (GO), Brazil.

2 Medicine student, Medical School, Federal University of Mato Grosso do Sul – UFMS, Campus Três Lagoas, Três Lagoas (MS), Brazil.

3 Medicine student, Medical School, Universitatea Ovidius din Constanța, Constanța, România.

4 Medicine student, Medical School, Federal University of Mato Grosso do Sul – UFMS, Campo Grande (MS), Brazil.

5 Medicine student, Medical School, University of Santo Amaro, São Paulo (SP), Brazil.

6 MD, PhD, Neurosurgeon, Post-Graduation in Teaching, PhD in Neurology, Neuroscience Center, Clínica Girassol, Luanda, Angola.

 

Received Aug 30, 2021

Accepted Jan 16, 2023