Camussoni, Francesco (2021) Aprendizaje automático y análisis estadístico para la evaluación de riesgo de rotura y comportamiento biomecánico de aneurismas intracraneales / Machine learning and statiscal analysis for rupture risk and biomechanical behaviour of intracraneal aneurysms. Proyecto Integrador Ingeniería Mecánica, Universidad Nacional de Cuyo, Instituto Balseiro.
![[img]](/style/images/fileicons/application_pdf.png)
| PDF (Tesis) Español 14Mb |
Resumen en español
Se estima que el 2 %-3% de de la población mundial presenta aneurismas intracraneales. Aunque los mismos suelen ser asintomáticos, la rotura y consecuente hemorragia del mismo, ponen en gran riesgo la vida de las personas, con tasas de mortalidad cercanas al 50 %. Junto a otros accidentes cerebro vasculares, representan la tercer causa de muerte en argentina. Sumado a esto, no existen al presente marcadores genéticos claros, ni factores de riesgo que in fluyan en el crecimiento y rotura de un aneurisma, existiendo variaciones significativas en la bibliografía. Entonces, existe el dilema de realizar una posible intervención quirúrgica innecesaria, de alta morbilidad, frente al riesgo que involucra la rotura del mismo. En el presente proyecto, se desarrollaron técnicas de aprendizaje automático para la clasificación del estado de rotura de aneurismas de la base de datos de Aneurisk. Para esto, se realizo una búsqueda exhaustiva de atributos relevantes al estado de rotura de aneurisma en antecedentes. Además, se desarrollaron técnicas de ingeniera de atributos como ser RFE y SPEC. Por otro lado, se utilizaron técnicas de reducción de dimensionalidad como ser UMAP, Isomap, PCA y el uso de un autoencoder. Finalmente, se generaron datos artificiales mediante la salida de un autoencoder. Los clasificadores propuestos fueron LR, KNN, SVC, XGB, GL y LGL, para los cuales se obtuvieron rendimientos máximos según AUC de 0.79, 0.79, 0.78, 0.8, 0.79, 0.82 para los clasificadores, respectivamente, para los datos de generalización. Por otro lado, se realizaron análisis estadísticos de los atributos obtenidos de un modelo biomecánico de laminas delgadas de Kircho-Love desarrollado en el proyecto PICTO-2016-0054, que enmarca el presente proyecto. Se encontró la dimensión intrínseca de los datos de salida, que resulto de 3, donde se utilizaron técnicas como PCA local, error de reconstrucción y la correlación de dimensión. En consiguiente, se realizo una reducción de dimensionalidad con Isomap a n de identificar propiedades de los atributos de salida. Se utilizo el aneurisma 34 como caso de estudio. En este, se idéntico que los efectos de pandeo están asociados a grandes curvaturas máximas, grandes relaciones de desplazamiento máximo y medio y grandes energías de membrana. Dichos resultados, fueron consistentes para los aneurismas 14, 42 y 90, como casos de control.
Resumen en inglés
It is estimated that 2% - 3% of the world population present intracranial aneurysms. Although they are usually asymptomatic, the rupture and its consequent hemorrhage,implies great risk for people's live with mortality rates close to 50 %. This illness, along with other cerebrovascular accidents, represent the third death cause in Argentina. In addition, there are no clear genetic markers present or risk factors that in fluence the growth and rupture of an aneurysm, with signicant variations in the literature. Then, there is the dilemma of a possible and unnecessary surgical intervention, with high morbidity, when an aneurysm is detected and the risk involved in its rupture. In the present project, machine learning techniques were developed for classifying the rupture status of aneurysms from the Aneurisk database. For this, an exhaustive search in antecedents was carried out, where relevant features to the state of rupture of the aneurysm were found. In addition, feature engineering techniques such as RFE and SPEC were developed. On the other hand, dimensionality reduction techniques were used, such as UMAP, Isomap, PCA and an autoencoder. Finally, articial data was generated by an autoencoder. The proposed classiers were LR, KNN, SVC, XGB, GL and LGL, where the obtained maximum performances according to AUC were 0.79, 0.79, 0.78, 0.8, 0.79, 0.82 for the classiers, respectively, for the generalization data. On the other hand, statistical analysis was carried out of the features obtained from a Kircho-Love thin-lamina biomechanical model developed in project PICTO-2016- 0054, which frames the present work. The intrinsic dimension of the output data was found, which was 3. For this, local PCA techniques, reconstruction error and dimension correlation were used. Consequently, dimensionality reduction techniques were performed with techniques such as Isomap and UMAP in order to identify properties of the output features. Aneurysm 34 was used as a case study. Buckling eeffcts were associated with large maximum curvatures, large maximum and mean displacement ratios and large membrane energies. These results were consistent for control aneurysms 14, 42 and 90.
| Tipo de objeto: | Tesis (Proyecto Integrador Ingeniería Mecánica) |
|---|---|
| Palabras Clave: | Machine learning; Aprendizaje automático; Statistics; Estadística; [Rupture risk; Riesgo de rotura; Biomechanics; Biomecánica; Intracraneal aneurysms; Aneurismas intracraneal] |
| Referencias: | [1] Python. http://www.python.org/. Accedido: 25 de junio de 2021. 2 [2] Numpy. https://numpy.org/. Accedido: 25 de junio de 2021. 2 [3] Pandas. https://pandas.pydata.org/. Accedido: 25 de junio de 2021. 3 [4] Matplotlib: Visualization with python. https://matplotlib.org/. Accedido:25 de junio de 2021. 3 [5] Plotly python open source graphing library. https://plotly.com/python/.Accedido: 25 de junio de 2021. 3 [6] Sickit-learn. https://scikit-learn.org/. Accedido: 25 de junio de 2021. 3, 26,28, 45 [7] Umap. https://umap-learn.readthedocs.io/. Accedido: 25 de junio de 2021.3, 42, 43, 61 [8] Keras. https://keras.io/. Accedido: 25 de junio de 2021. 3 [9] Xgboost. https://xgboost.readthedocs.io/. Accedido: 25 de junio de 2021.3, 34 [10] Group lasso. https://group-lasso.readthedocs.io/en/latest/. Accedido:25 de junio de 2021. 3, 35 [11] Rinkel, G. J. E., Djibuti, M., Algra, A., van Gijn, J. Prevalence and risk of ruptureof intracranial aneurysms. Stroke, 29 (1), 251-256, ene. 1998. URL https://doi.org/10.1161/01.str.29.1.251. 5, 6 [12] Juvela, S., Poussa, K., Lehto, H., Porras, M. Natural history of unruptured intracranial aneurysms. Stroke, 44 (9), 2414-2421, sep. 2013. URL https://doi.org/10.1161/strokeaha.113.001838. 6 [13] Zacharia, B. E., Hickman, Z. L., Grobelny, B. T., DeRosa, P., Kotchetkov, I.,Ducruet, A. F., et al. Epidemiology of aneurysmal subarachnoid hemorrhage. Neurosurgery Clinics of North America, 21 (2), 221-233, abr. 2010. URL https://doi.org/10.1016/j.nec.2009.10.002. 6 [14] Connolly, E. S., Rabinstein, A. A., Carhuapoma, J. R., Derdeyn, C. P., Dion, J., Higashida, R. T., et al. Guidelines for the management of aneurysmal subarachnoid hemorrhage. Stroke, 43 (6), 1711-1737, jun. 2012. URL https://doi.org/10.1161/str.0b013e3182587839. 6 [15] Naggara, O. N., Lecler, A., Oppenheim, C., Meder, J.-F., Raymond, J. Endovascular treatment of intracranial unruptured aneurysms: A systematic review of the literature on safety with emphasis on subgroup analyses. Radiology, 263 (3), 828-835, jun. 2012. URL https://doi.org/10.1148/radiol.12112114. 6 [16] Giaroli, E., Moyano, L. G., Petra, E., Lagos, M., Curiale, A. H., Millan, D. Morphometric descriptors of intracranial sacular aneurysms: principal component analysis. En: Congreso Latinoamericano de Ingeniera y Ciencias Aplicadas (CLICAP), pags. 619{626. Formacion universitaria, 2018. 6, 8, 17 [17] Piccinelli, M., Steinman, D. A., Hoi, Y., Tong, F., Veneziani, A., Antiga, L. Automatic neck plane detection and 3d geometric characterization of aneurysmal sacs. Annals of Biomedical Engineering, 40 (10), 2188-2211, abr. 2012. URL https://doi.org/10.1007/s10439-012-0577-5. 7 [18] Chollet, F. Deep Learning with Python. Manning, 2018. 7, 8, 19, 36, 38 [19] Geron, A. Hands-On Machine Learning with Scikit-Learn, Keras, and Tensor flow. O'Reilly, 2019. 7, 19, 31, 32, 33, 39, 40 [20] Aneurisk. http://ecm2.mathcs.emory.edu/aneuriskweb/index. Accedido: 25 de junio de 2021. 1, 9, 10, 11, 47, 52, 53, 55, 75, 79, 103 [21] Zanaty, M., Chalouhi, N., Tjoumakaris, S. I., Gonzalez, L. F., Rosenwasser, R. H., Jabbour, P. M. Aneurysm geometry in predicting the risk of rupture. a review of the literature. Neurological Research, 36 (4), 308-313, feb. 2014. URL https://doi.org/10.1179/1743132814y.0000000327. 14, 16, 17, 64 [22] Chien, A., Callender, R. A., Yokota, H., Salamon, N., Colby, G. P., Wang, A. C., et al. Unruptured intracranial aneurysm growth trajectory: occurrence and rate of enlargement in 520 longitudinally followed cases. Journal of Neurosurgery, 132 (4), 1077-1087, abr. 2020. URL https://doi.org/10.3171/2018.11.jns181814. 14, 18 [23] Detmer, F. J., Luckehe, D., Mut, F., Slawski, M., Hirsch, S., Bijlenga, P., et al. Comparison of statistical learning approaches for cerebral aneurysm rupture assessment. International Journal of Computer Assisted Radiology and Surgery, 15 (1), 141-150, sep. 2019. URL https://doi.org/10.1007/s11548-019-020 65-2. 14, 17, 18, 64, 65 [24] Raghavan, M. L., Ma, B., Harbaugh, R. E. Quantied aneurysm shape and rupture risk. Journal of Neurosurgery, 102 (2), 355-362, feb. 2005. URL https://doi.org/10.3171/jns.2005.102.2.0355. 16 [25] Weir, B., Disney, L., Karrison, T. Sizes of ruptured and unruptured aneurysms in relation to their sites and the ages of patients. Journal of Neurosurgery, 96 (1), 64-70, ene. 2002. URL https://doi.org/10.3171/jns.2002.96.1.0064. 16 [26] Szikora, I., Paal, G., Ugron, A., Nasztanovics, F., Marosfoi, M., Berentei, Z., et al. Impact of aneurysmal geometry on intraaneurysmal flow: a computerized flow simulation study. Neuroradiology, 50 (5), 411-421, ene. 2008. URL https://doi.org/10.1007/s00234-007-0350-x. 16 [27] Ryu, C.-W., Kwon, O.-K., Koh, J. S., Kim, E. J. Analysis of aneurysm rupture in relation to the geometric indices: aspect ratio, volume, and volume-to-neck ratio. Neuroradiology, 53 (11), 883-889, nov. 2010. URL https://doi.org/10.1007/s00234-010-0804-4. 16, 17, 55 [28] Chien, A., Sayre, J., Viñuela, F. Comparative morphological analysis of the geometry of ruptured and unruptured aneurysms. Neurosurgery, 69 (2), 349-356, mar. 2011. URL https://doi.org/10.1227/neu.0b013e31821661c3. 16 [29] Matsukawa, H., Uemura, A., Fujii, M., Kamo, M., Takahashi, O., Sumiyoshi, S. Morphological and clinical risk factors for the rupture of anterior communicating artery aneurysms. Journal of Neurosurgery, 118 (5), 978-983, mayo 2013. URL https://doi.org/10.3171/2012.11.jns121210. 17, 55 [30] Dhar, S., Tremmel, M., Mocco, J., Kim, M., Yamamoto, J., Siddiqui, A. H., et al. MORPHOLOGY PARAMETERS FOR INTRACRANIAL ANEURYSM RUPTURE RISK ASSESSMENT. Neurosurgery, 63 (2), 185-197, ago. 2008. URL https://doi.org/10.1227/01.neu.0000316847.64140.81. 17 [31] Andreas Muller, S. G. Introduction to Machine Learning with Python. O'Reilly, 2017. 19, 21, 28, 29, 34 [32] Machine learning mastery - recursive feature elimination. https://machinelea rningmastery.com/rfe-feature-selection-in-python/. Accedido: 25 de junio de 2021. 23 [33] Zhao, Z., Liu, H. Spectral feature selection for supervised and unsupervised learning, 2007. URL https://doi.org/10.1145/1273496.1273641. 24 [34] McCulloch, W. S., Pitts, W. A logical calculus of the ideas immanent in nervous activity. The Bulletin of Mathematical Biophysics, 5 (4), 115-133, dic. 1943. URL https://doi.org/10.1007/bf02478259. 36 [35] Rosenblatt, F. The perceptron: A probabilistic model for information storage and organization in the brain. Psychological Review, 65 (6), 386-408, 1958. URL https://doi.org/10.1037/h0042519. 36 [36] Pearson, K. LIII. on lines and planes of closest t to systems of points in space. The London, Edinburgh, and Dublin Philosophical Magazine and Journal of Science, 2 (11), 559-572, nov. 1901. URL https://doi.org/10.1080/14786440109462 720. 40 [37] Millan, D., Arroyo, M. Nonlinear manifold learning for model reduction in nite elastodynamics. Computer Methods in Applied Mechanics and Engineering, 261- 262, 118-131, jul. 2013. URL https://doi.org/10.1016/j.cma.2013.04.007. 40 [38] Isomap. https://en.wikipedia.org/wiki/Isomap. Accedido: 25 de junio de 2021. 41 [39] Floyd-warshall algorithm. https://en.wikipedia.org/wiki/Floyd%E2%80%93 Warshall algorithm. Accedido: 25 de junio de 2021. 42 [40] Multidimensional scaling. https://en.wikipedia.org/wiki/Multidimension al scaling. Accedido: 25 de junio de 2021. 42 [41] Curva roc. https://es.wikipedia.org/wiki/Curva ROC. Accedido: 25 de junio de 2021. 46 [42] Multicollinearity. https://en.wikipedia.org/wiki/Multicollinearity#Con sequences of multicollinearity. Accedido: 25 de junio de 2021. 48 [43] How to get the most contributing feature in any classier sklearn. https://stac koverflow.com/questions/42088336/how-to-get-the-most-contributing- feature-in-any-classifier-sklearn-for-example-d. Accedido: 25 de junio de 2021. 54 [44] Muzi, N., Schoeman, P., Ferrari, I., Millan, D. Spretratamiento de modelos geométricos para el calculo hiperelástico de aneurismas cerebrales mediante modelos de laminas delgadas. 2010. 73, 75, 76 [45] Ferrari, I., Muzi, N., Schoeman, P., Moyano, L., Millan, D. Análisis computacional de la biomecánica de aneurismas cerebrales mediante un modelo de laminas delgadas de kircho-love con espesor variable. 2020. 73, 74, 77 |
| Materias: | Medicina > Medicina informática |
| Divisiones: | Gcia. de área de Investigación y aplicaciones no nucleares > Gcia. de Física > Sistemas complejos y altas energías > Física estadística interdisciplinaria |
| Código ID: | 1018 |
| Depositado Por: | Tamara Cárcamo |
| Depositado En: | 06 May 2022 15:17 |
| Última Modificación: | 06 May 2022 15:17 |
Personal del repositorio solamente: página de control del documento
