Deep learning aplicado para la detección de hemorragias y tumores cerebrales
DOI:
https://doi.org/10.5380/atoz.v10i3.81284Keywords:
Deep Learning, Redes convolucionales, Aplicaciones prácticas, Bioinformática, Tumores cerebralesAbstract
Introducción: Un problema que afecta a la salud en Chile se refiere a las patologías cerebrales, toma de exámenes y el alto tiempo de espera para la obtención de los resultados (retrasando el diagnóstico y tratamiento). Actualmente, los exámenes se envían al extranjero para ser procesados y el tiempo de espera juega en contra del paciente. Dada esta realidad, nuestro documento propone un modelo de deep learning para la predicción de imágenes cerebrales que permita obtener un diagnóstico previo, pero no definitivo, en virtud de disminuir el tiempo del proceso y, de ser necesario, priorizar a los pacientes cuya vida estaría potencialmente en riesgo. Métodos: El desarrollo utilizó un enfoque RAD iterativo y las imágenes se recogieron de Kaggle. Adicionalmente, se redimensiona el dataset para normalizar el tamaño y generamos nuevas imágenes utilizando “data augmentation”. Las imágenes fueron procesadas en redes convolucionales, indagando en distintas configuraciones para la red, su optimizador y la función de activación, hasta llegar a un modelo que consideramos razonable. Resultados: Con el modelo definitivo, los resultados superan el 80% de precisión en las predicciones y descubrimos que separar patologías (hemorragias y tumores) fue crucial para este resultado. Conclusiones: Hemos logrado una herramienta de diagnóstico previo, pero se debe continuar la investigación en virtud de aumentar la precisión. Un próximo paso considera ampliar el dataset con imágenes de otras fuentes y separar el modelo para analizar patologías de forma independiente. Motivamos a seguir investigando ya que este tipo de apoyo puede contribuir a salvar vidas.
References
Aggarwal, C. (2018). Neural Networks and Deep Learning a Textbook. Yorktown Heights, USA; Springer International Publishing AG.
Ari, A. & Hanbay, D. (2018). Deep learning based brain tumor classification and detection system. Turkish journal of Electrical Engineering & Computer Sciences. 26. 2275-2286. doi: http://dx.doi.org/10.3906/elk-1801-8.
Chollet, F. (2017). Deep Learning with Python. Shelter Island, New York, USA; Manning Publications.
Gholamalinejad, H. & Khosravi, H. (2020). Pooling Methods in Deep Neural Networks, a Review. Computing Research Repository (CoRR). Recuperado de https://arxiv.org/abs/2009.07485.
Glorot, X., Bordes, A. & Bengio, Y. (2010). Deep Sparse Rectifier Neural Networks. Journal of Machine Learning Research. 15. Recuperado de https://proceedings.mlr.press/v15/glorot11a/glorot11a.pdf
Hassan, M. (2018). VGG16 – Convolutional Network for Classification and Detection. Recuperado de https://neurohive.io/en/popular-networks/vgg16/
Knoll, F. et al. (2020). Deep-Learning Methods for Parallel Magnetic Resonance Imaging Reconstruction: A Survey of the Current Approaches, Trends, and Issues. IEEE Signal Processing Magazine, 37(1), 128-140. doi: https://doi.org/10.1109/MSP.2019.2950640
Lecun, Y. (1989). Generalization and network design strategies. Zurich, Switzerland: Elsevier.
McCulloch, W.S. & Pitts, W. (1943). A logical calculus of the ideas immanent in nervous activity. Bulletin of Mathematical Biophysics. 5, 115–133. doi: https://doi.org/10.1007/BF02478259
O'Shea, K., & Nash, R. (2015). An Introduction to Convolutional Neural Networks. Computing Research Repository (CoRR). Recuperado de https://arxiv.org/abs/1511.08458
S. J. Pan, S. J., & Yang, Q. (2010). A Survey on Transfer Learning. IEEE Transactions on Knowledge and Data Engineering. 22(10), 1345-1359. doi: https://doi.org/10.1109/TKDE.2009.191
Petersen, P. et al. (2020). Topological properties of the set of functions generated by neural networks of fixed size. Computing Research Repository (CoRR). Recuperado de https://arxiv.org/abs/1806.08459
Petersen, P., & Voigtländer, F. (2017). Optimal approximation of piecewise smooth functions using deep ReLU neural networks. Neural networks: the official journal of the International Neural Network Society. doi: http://dx.doi.org/10.1016/j.neunet.2018.08.019
Sarkar, D. & Bali, R. (2018). Hands-On Transfer Learning with Python. Shelter Island, New York, USA; Manning Publications.
Schmidhuber, J. (2014). Deep Learning in Neural Networks: An Overview. Computing Research Repository (CoRR). Recuperado de https://arxiv.org/abs/1404.7828
Simonyan, K. & Zisserman, A. (2015). Very Deep Convolutional Networks for large-scale image recognition. Computing Research Repository (CoRR). Recuperado de https://arxiv.org/abs/1409.1556
Sorias O. et al. (2009). Handbook of Research On Machine Learning Applications and Trends: Algorithms, Methods and Techniques. Hershey, PA, USA; Information Science Reference - Imprint of: IGI Publishing.
Zhantao, C. et al. (2017). Breast Tumor Detection in Ultrasound Images Using Deep Learning. Lecture Notes in Computer Science, 10530. Springer, Cham. doi: https://doi.org/10.1007/978-3-319-67434-6_14
Downloads
Published
How to Cite
Issue
Section
License
Atoz is a open access journal and the authors have permission and are encouraged to deposit their papers in personal web pages, institutional repositories or portals before (pre-print) or after (post-print) the publication at AtoZ. It is just asked, when and where possible, the mention, as a bibliographic reference (including the atributted URL), to the AtoZ Journal.
The authors license the AtoZ for the solely purpose of disseminate the published work (peer reviewed version/post-print) in aggregation, curation and indexing systems.
The AtoZ is a Diadorim/IBICT green academic journal.
All the journal content (including instructions, editorial policies and templates) - except where otherwise indicated - is under a Creative Commons Attribution 4.0 International, since October 2020.
When published by this journal, articles are free to share (copy and redistribute the material in any support or format for any purpose, even commercial) and adapt (remix, transform, and create from the material for any purpose , even if commercial). You must give appropriate credit , provide a link to the license, and indicate if changes were made
AtoZ does not apply any charges regarding manuscripts submission/processing and papers publication.
























