Sistema web de reconocimiento y clasificación de patologías a través de imágenes médicas basado en técnicas de aprendizaje de máquina

La detección de cáncer de tiroides es un proceso que en la actualidad se realiza mediante la interpretación manual que realizan radiólogos especialistas, estas se clasifican utilizando una prueba de tamizaje (discriminatoria) conocida como EU- TIRADS 2017 [2], que determina el grado de malignidad de...

Full description

Autores:
Arias Trillos, Yhary Estefanía
Tipo de recurso:
Trabajo de grado de pregrado
Fecha de publicación:
2019
Institución:
Universidad Autónoma de Bucaramanga - UNAB
Repositorio:
Repositorio UNAB
Idioma:
spa
OAI Identifier:
oai:repository.unab.edu.co:20.500.12749/7053
Acceso en línea:
http://hdl.handle.net/20.500.12749/7053
Palabra clave:
Systems engineer
Technological innovations
Data set
Deep learning
Maching
Data increase
Neural networks
Artificial intelligence
Ultrasound
Radiology
Area under the curve
Confusion matrix
Concordance study
Cancer diagnosis
X-rays
Medical examinations
Diagnostic service
Endocrine glands
Ingeniería de sistemas
Innovaciones tecnológicas
Inteligencia artificial
Cáncer diagnóstico
Rayos x
Exámenes médicos
Servicio de diagnóstico
Glándulas endocrinas
Conjunto de datos
Aprendizaje profundo
Aumento de datos
Redes neuronales
Ultrasonido
Radiología
Área bajo la curva
Matriz de confusión
Estudio de concordancia
Rights
openAccess
License
http://creativecommons.org/licenses/by-nc-nd/2.5/co/
id UNAB2_dd66118d356dec40db7f05ae070bfd38
oai_identifier_str oai:repository.unab.edu.co:20.500.12749/7053
network_acronym_str UNAB2
network_name_str Repositorio UNAB
repository_id_str
dc.title.spa.fl_str_mv Sistema web de reconocimiento y clasificación de patologías a través de imágenes médicas basado en técnicas de aprendizaje de máquina
dc.title.translated.eng.fl_str_mv Web recognition and classification system pathologies through medical images based on machine learning techniques
title Sistema web de reconocimiento y clasificación de patologías a través de imágenes médicas basado en técnicas de aprendizaje de máquina
spellingShingle Sistema web de reconocimiento y clasificación de patologías a través de imágenes médicas basado en técnicas de aprendizaje de máquina
Systems engineer
Technological innovations
Data set
Deep learning
Maching
Data increase
Neural networks
Artificial intelligence
Ultrasound
Radiology
Area under the curve
Confusion matrix
Concordance study
Cancer diagnosis
X-rays
Medical examinations
Diagnostic service
Endocrine glands
Ingeniería de sistemas
Innovaciones tecnológicas
Inteligencia artificial
Cáncer diagnóstico
Rayos x
Exámenes médicos
Servicio de diagnóstico
Glándulas endocrinas
Conjunto de datos
Aprendizaje profundo
Aumento de datos
Redes neuronales
Ultrasonido
Radiología
Área bajo la curva
Matriz de confusión
Estudio de concordancia
title_short Sistema web de reconocimiento y clasificación de patologías a través de imágenes médicas basado en técnicas de aprendizaje de máquina
title_full Sistema web de reconocimiento y clasificación de patologías a través de imágenes médicas basado en técnicas de aprendizaje de máquina
title_fullStr Sistema web de reconocimiento y clasificación de patologías a través de imágenes médicas basado en técnicas de aprendizaje de máquina
title_full_unstemmed Sistema web de reconocimiento y clasificación de patologías a través de imágenes médicas basado en técnicas de aprendizaje de máquina
title_sort Sistema web de reconocimiento y clasificación de patologías a través de imágenes médicas basado en técnicas de aprendizaje de máquina
dc.creator.fl_str_mv Arias Trillos, Yhary Estefanía
dc.contributor.advisor.spa.fl_str_mv Ortiz Beltrán, Ariel
dc.contributor.author.spa.fl_str_mv Arias Trillos, Yhary Estefanía
dc.contributor.cvlac.*.fl_str_mv Ortiz Beltrán, Ariel [0001459925]
dc.contributor.researchgate.*.fl_str_mv Ortiz Beltrán, Ariel [Ariel-Ortiz-Beltran]
dc.contributor.researchgroup.spa.fl_str_mv Grupo de Investigación Tecnologías de Información - GTI
Grupo de Investigaciones Clínicas
dc.subject.keywords.eng.fl_str_mv Systems engineer
Technological innovations
Data set
Deep learning
Maching
Data increase
Neural networks
Artificial intelligence
Ultrasound
Radiology
Area under the curve
Confusion matrix
Concordance study
Cancer diagnosis
X-rays
Medical examinations
Diagnostic service
Endocrine glands
topic Systems engineer
Technological innovations
Data set
Deep learning
Maching
Data increase
Neural networks
Artificial intelligence
Ultrasound
Radiology
Area under the curve
Confusion matrix
Concordance study
Cancer diagnosis
X-rays
Medical examinations
Diagnostic service
Endocrine glands
Ingeniería de sistemas
Innovaciones tecnológicas
Inteligencia artificial
Cáncer diagnóstico
Rayos x
Exámenes médicos
Servicio de diagnóstico
Glándulas endocrinas
Conjunto de datos
Aprendizaje profundo
Aumento de datos
Redes neuronales
Ultrasonido
Radiología
Área bajo la curva
Matriz de confusión
Estudio de concordancia
dc.subject.lemb.spa.fl_str_mv Ingeniería de sistemas
Innovaciones tecnológicas
Inteligencia artificial
Cáncer diagnóstico
Rayos x
Exámenes médicos
Servicio de diagnóstico
Glándulas endocrinas
dc.subject.proposal.spa.fl_str_mv Conjunto de datos
Aprendizaje profundo
Aumento de datos
Redes neuronales
Ultrasonido
Radiología
Área bajo la curva
Matriz de confusión
Estudio de concordancia
description La detección de cáncer de tiroides es un proceso que en la actualidad se realiza mediante la interpretación manual que realizan radiólogos especialistas, estas se clasifican utilizando una prueba de tamizaje (discriminatoria) conocida como EU- TIRADS 2017 [2], que determina el grado de malignidad del nódulo tiroideo. La escasez de profesionales y la creciente demanda de este tipo de estudios plantea el problema de la automatización a través de algoritmos de aprendizaje de máquina como los basados en Deep Learning y específicamente, las Redes Neuronales Convolucionales, que han sido probadas anteriormente con éxito para la clasificación de otro tipo de imágenes médicas. En un trabajo anterior, con un dataset de 2000 imágenes balanceado entre 4 categorías (TI-RADS2 - TI-RADS5) se logró una medida de precisión (accuracy) cercana del 65% y una pérdida logarítmica (cross-entropy loss) cercana a 0.78. Sin embargo, este artículo plantea el estudio exploratorio para una posible optimización del algoritmo a través de diferentes pruebas medibles en su parametrización. Las variables que serán ajustadas son: El número de capas convolucionales, el tamaño de la máscara de convolución, las funciones de activación, el número de neuronas en la capa densa, el uso de más capas densas para el aprendizaje, el uso de dropouts aleatorios para controlar el sobreajuste (overfitting), entre otros. La medición comparativa se realiza a través de los valores de precisión, pérdida, la matriz de confusión, y el área bajo la curva ROC. Al final del documento se describe la mejor combinación de los parámetros evaluados y las observaciones pertinentes.
publishDate 2019
dc.date.issued.none.fl_str_mv 2019
dc.date.accessioned.none.fl_str_mv 2020-07-27T19:19:09Z
dc.date.available.none.fl_str_mv 2020-07-27T19:19:09Z
dc.type.driver.none.fl_str_mv info:eu-repo/semantics/bachelorThesis
dc.type.local.spa.fl_str_mv Trabajo de Grado
dc.type.coar.none.fl_str_mv http://purl.org/coar/resource_type/c_7a1f
dc.type.redcol.none.fl_str_mv http://purl.org/redcol/resource_type/TP
format http://purl.org/coar/resource_type/c_7a1f
dc.identifier.uri.none.fl_str_mv http://hdl.handle.net/20.500.12749/7053
dc.identifier.instname.spa.fl_str_mv instname:Universidad Autónoma de Bucaramanga - UNAB
dc.identifier.reponame.spa.fl_str_mv reponame:Repositorio Institucional UNAB
dc.identifier.repourl.spa.fl_str_mv repourl:https://repository.unab.edu.co
url http://hdl.handle.net/20.500.12749/7053
identifier_str_mv instname:Universidad Autónoma de Bucaramanga - UNAB
reponame:Repositorio Institucional UNAB
repourl:https://repository.unab.edu.co
dc.language.iso.spa.fl_str_mv spa
language spa
dc.relation.references.spa.fl_str_mv [1] Chapter 1: Supervised Learning and Naive Bayes Classification — Part 2 (Coding). (2017). Retrieved from https://medium.com/machine-learning101/chapter-1supervised-learning-and-naive-bayes-classification-part-2coding-5966f25f1475
[2] clinicalkey.es. (2018). Thyroid Imaging. clinicalkey. Retrieved from https://wwwclinicalkey-es.aure.unab.edu.co/service/content/pdf/watermarked/3s2.0B9780323189071000792.pdf?locale=es_ES
[3] clinicalkey.es. (2018). Thyroid Neoplasia. Retrieved from https://wwwclinicalkeyes.aure.unab.edu.co/service/content/pdf/watermarked/3s2.0B9780323189071000925.pdf?locale=es_ES
[4] ClinicalKey.es. (2018). The Thyroid Gland. Estados Unidos. Retrieved from https://www-clinicalkey- es.aure.unab.edu.co/service/content/pdf/watermarked/3- s2.0B9780323401715000195.pdf?locale=es_ES
[5] Department of Computer Science. (1999). Correlation-based Feature Selection for Machine Learning. Hamilton, NewZealand. Retrieved from https://www.lri.fr/~pierres/donn%E9es/save/these/articles/lprqueue/hall99correlationb ased.pdf
[6] Department of Computer Science University of Waikato. (1999). Feature Selection for Machine Learning: Comparing a Correlation-based Filter Approach to the Wrapper. New Zealand: waikato. Retrieved from http://www.aaai.org/Papers/FLAIRS/1999/FLAIRS99-042.pdf
[7] ESMERILADO - Definición - Significado. (2018). Retrieved from https://diccionario.motorgiga.com/esmerilado
[8] Exámenes PET CT y RM: ¿Qué son y cómo usarlos en la medicina diagnóstica?. (2015). Retrieved from http://www.mv.com.br/es/blog/examenes- -pet-ct-y-rm---que-son-y-como-usarlos-en-la-medicina-diagnosticar
[9] Medicina nuclear: SPECT y PET en tumores primarios del Sistema Nervioso | NeuroWikia. (2010). Retrieved from http://www.neurowikia.es/content/medicina-nuclear-spect-y-pet-en-tumoresprimariosdel-sn
[10] Northside Radiology Associates. (2016). MIBG scintiscan. Atlanta: Editorial team. Retrieved from https://medlineplus.gov/ency/article/003830.htm
[11] Positron. (2018). Retrieved from https://en.wikipedia.org/wiki/Positron
[12] Pruebas para detectar el cáncer de tiroides. (2016). Retrieved from https://www.cancer.org/es/cancer/cancer-de-tiroides/detecciondiagnosticoclasificacion-por-etapas/como-se-diagnostica.html
[13] University of Washington School of Medicine. (2016). Gammagrafía de la tiroides. Seattle: Editorial Director y A.D.A.M. Editorial team. Retrieved from https://medlineplus.gov/spanish/ency/article/003829.htm
[14] working paper series. (2000). Correlation - based feature selection ffor discrete and numeric class Machine Learning. New Zealand. Retrieved from https://researchcommons.waikato.ac.nz/bitstream/handle/10289/1024/uow-cswp- 2000-08.pdf?sequence=1&isAllowed=y
[15] Yodo-131. (2018). Retrieved from https://es.wikipedia.org/wiki/Yodo-131
[16] ¿Qué es una API REST? - Idento. (2018). Retrieved from https://www.idento.es/blog/desarrollo-web/que-es-una-api-rest/
[17] Cloud AutoML - Custom Machine Learning Models | AutoML | Google Cloud. (2018). Retrieved from https://cloud.google.com/automl/
[18] Convolutional Neural Networks for Visual Recognition. (2018). Retrieved from http://cs231n.github.io/convolutional-networks/
[19] general, M. (2018). Salarios de Médico/a general en Colombia | Indeed.com. Retrieved from https://co.indeed.com/salaries/M%C3%A9dico/a-generalSalaries
[20] Journal of Physics: Conference Series. (2016). Recent development of feature extraction and classification multispectral/hyperspectral images: a systematic literature review. Retrieved from http://iopscience.iop.org/article/10.1088/17426596/801/1/012045/pdf
[21] Judith Marcin, M. (2018). MRI Scans: Definition, uses, and procedure. Retrieved from https://www.medicalnewstoday.com/articles/146309.php
[22] Judith Marcin, M. (2018). MRI Scans: Definition, uses, and procedure. Retrieved from https://www.medicalnewstoday.com/articles/146309.php
[21] Principal Component Analysis [PCA]. (2017). Retrieved from https://medium.com/100-days-of-algorithms/day-92-pca-bdb66840a8fb
[22] ProClass Software - 2018 Reviews, Pricing & Demo. (2018). Retrieved from https://www.softwareadvice.com/registration/proclass-profile/
[23] ProClass Software - 2018 Reviews, Pricing & Demo. (2018). Retrieved from https://www.softwareadvice.com/registration/proclass-profile/
[24] Purves, D., Augustine, G., Fitzpatrick, D., Katz, L., LaMantia, A., McNamara, J., & Williams, S. (2018). Types of Eye Movements and Their Functions. Retrieved from https://www.ncbi.nlm.nih.gov/books/NBK10991/
[25] Robust image hashing using ring partition-PGNMF and local features. (2016). Retrieved from https://www.ncbi.nlm.nih.gov/pmc/articles/PMC5118381/
[26] School of Electronics and Control Engineering Chang'an. (2012). Hyperspectral Image Classification Based on Spectral-Spatial Feature Extraction. Xi’an, China. Retrieved from https://ieeexplore- ieeeorg.aure.unab.edu.co/stamp/stamp.jsp?tp=&arnumber=7958808&tag=1
[27] sistemas, I. (2018). Salarios de Ingeniero/a en sistemas en Colombia | Indeed.com. Retrieved from https://co.indeed.com/salaries/Ingeniero/a-ensistemas-Salaries
[26] What is a data set? - Quora. (2017). Retrieved from https://www.quora.com/What-is-adata-set
[27] Brownlee, J. (2016). What is Deep Learning? Retrieved from https://machinelearningmastery.com/what-is-deep-learning
[27] Brownlee, J. (2016). What is Deep Learning? Retrieved from https://machinelearningmastery.com/what-is-deep-learning
[29] Raicea, R. (2018). Want to know how Deep Learning works? Here’s a quick guide for everyone. Retrieved from https://medium.freecodecamp.org/want-toknow-how-deeplearning-works-heres-a-quick-guide-for-everyone1aedeca88076
[30] Tiempo, C. (2018). Sueldo de un profesional con posgrado. Retrieved from https://www.portafolio.co/economia/empleo/un-trabajador-con-posgrado-ganaenpromedio-3-3-millones-mas-que-un-bachiller-512462
[31] School of Electronics and Control Engineering Chang'an University (2017). Hyperspectral Image Classification Based on SpectralSpatial Feature Extraction. Xi’an, China. Retrieved from https://ieeexplore-ieeorg.aure.unab.edu.co/stamp/stamp.jsp?tp=&arnumber=7958808
[32] School of Electronics and Control Engineering. (2013). Using Nonnegative Matrix Factorization with Projected Gradient for Hyperspectral Images Feature Extraction. Retrieved from https://ieeexplore-ieee- org.aure.unab.edu.co/stamp/stamp.jsp?tp=&arnumber=6566423&ta g=1
[33] Dharmsinh Desai University. (2009). Project Classification Using Soft Computing. Nadiad, Gujarat. Retrieved from https://ieeexplore- ieeeorg.aure.unab.edu.co/stamp/stamp.jsp?tp=&arnumber=5376511&ta g=1
[34] Dharmsinh Desai University. (2009). Project Classification Using Soft Computing. Nadiad, Gujarat. Retrieved from https://ieeexplore- ieeeorg.aure.unab.edu.co/stamp/stamp.jsp?tp=&arnumber=5376511&ta g=1
[35] Department of Computer Science California State Polytechnic University. (2016). Oculomotor Plant Feature Extraction from Human Saccadic Eye Movements. Pomona, USA. Retrieved from https://ieeexplore-ieee- org.aure.unab.edu.co/stamp/stamp.jsp?tp=&arnumber=8424698
[36] Bharathidasan College of Arts & Science. (2016). Fusion of Big Data and Neural Networks for Predicting Thyroid. Tamilnadu, India. Retrieved from https://ieeexplore-ieee- org.aure.unab.edu.co/stamp/stamp.jsp?tp=&arnumber=7955223
[36] Bharathidasan College of Arts & Science. (2016). Fusion of Big Data and Neural Networks for Predicting Thyroid. Tamilnadu, India. Retrieved from https://ieeexplore-ieee- org.aure.unab.edu.co/stamp/stamp.jsp?tp=&arnumber=7955223
[37] Bernard, O., Lalande, A., Zotti, C., & Cervenansky, F. (2018). Deep Learning Techniques for Automatic MRI Cardiac Multi-structures Segmentation and Diagnosis: Is the Problem Solved? Retrieved from https://ieeexplore-ieee- org.aure.unab.edu.co/stamp/stamp.jsp?tp=&arnumber=8360453
[38] School of Informatics and Computing, Indiana University. (2015). Temporal Pattern and Association Discovery of Diagnosis Codes using Deep Learning. IN, USA. Retrieved from https://ieeexploreieee- org.aure.unab.edu.co/stamp/stamp.jsp?tp=&arnumber=7349719&ta g=1
[39] Amazon Rekognition – Videos e imágenes – AWS. (2018). Retrieved from https://aws.amazon.com/es/rekognition/?hp=tile&soexp=below
[40] AMI de aprendizaje profundo de Amazon. (2018). Retrieved from https://aws.amazon.com/es/machine-learning/amis/?hp=tile&soexp=below
[41] Amazon Polly. (2018). Retrieved from https://aws.amazon.com/es/polly/?hp=tile&so-exp=below pruébelo, V. (2018). Chatbot | Deep learning | Amazon Lex. Retrieved from https://aws.amazon.com/es/lex/?hp=tile&soexp=below
[42] Amazon Translate – Traducción de máquina neural – AWS. (2018). Retrieved from https://aws.amazon.com/es/translate/?hp=tile&soexp=below
[43] What is Watson. (2018). Retrieved from https://www.ibm.com/watson/about/index.html ©2018 IEEE. (2018). Delta Univ. for Science and Technology. Gamasa City, Egypt, from https://ieeexplore-ieee- org.aure.unab.edu.co/stamp/stamp.jsp?tp=&arnumber=8358209
[44] National Chung Hsing University. (2009). Optimal Grouping by using Genetic Algorithm and Support Vector Machines. Retrieved from https://ieeexplore-ieee- org.aure.unab.edu.co/stamp/stamp.jsp?tp=&arnumber=5420079
[45] EEE TRANSACTIONS ON BIOMEDICAL ENGINEERING. (2010). Thyroid Segmentation and Volume Estimation in Ultrasound Images. Retrieved from https://ieeexplore-ieee- org.aure.unab.edu.co/stamp/stamp.jsp tp=&arnumber=5415666&tag=1
[46] IEEE. (2017). Tiroid Kanserinde BilgisayarlÕ Tomografi Temelli Yeni Öznitelikler Computerized Tomography Based Novel Features in Thyroid Cancer. Retrieved from https://ieeexplore-ieee- org.aure.unab.edu.co/stamp/stamp.jsp?tp=&arnumber=8238050
[47] IEEE. (2009). Computer-Aided Diagnosis of Thyroid Malignancy Using an Artificial Immune System Classification Algorithm. Retrieved from https://ieeexplore-ieee- org.aure.unab.edu.co/stamp/stamp.jsp?tp=&arnumber=4539694
[48] AIMBE. (2012). Automated Benign & Malignant Thyroid Lesion Characterization and Classification in 3D Contrast-Enhanced Ultrasound. IEEE. Retrieved from https://ieeexplore-ieee- org.aure.unab.edu.co/stamp/stamp.jsp?tp=&arnumber=6345965
[49] The department of thyroid surgery. (2012). Discussion about misdiagnosed reasons and reoperation of thyroid cancer. Changchun city, China. Retrieved from https://ieeexplore- ieeeorg.aure.unab.edu.co/stamp/stamp.jsp?tp=&arnumber=6291383
[50] Department of Electronics and Computer Engineering, Gifu University Yanagido. (1992). Neural Network Approach for the ComputerAided Diagnosis of Coronary Artery Diseases in Nuclear Medicine. Retrieved from https://ieeexplore-ieeeorg.aure.unab.edu.co/stamp/stamp.jsp?tp=&arnumber=227168
[51] Shandong Provincial Key Laboratory of Computer Networks. (2016). Web Identification Image Recognition Based on Deep Learning. Jinan, China. Retrieved from https://ieeexplore-ieeeorg.aure.unab.edu.co/stamp/stamp.jsp?tp=&arnumber=7726261
[52] College of Computer Science and Technology. (2012). Interface Schema Matching with the Machine Learning for Deep Web. Harbin, P.R. China. Retrieved from https://ieeexplore-ieee.org.aure.unab.edu.co/stamp/stamp.jsp?tp=&arnumber=6526056
[53] Towards Data Science. (2019). Metrics to Evaluate your Machine Learning Algorithm. [online] Available at: https://towardsdatascience.com/metrics-to-evaluate-your-machinelearning-algorithm-f10ba6e38234 [Accessed 5 Jun. 2019].
dc.rights.uri.*.fl_str_mv http://creativecommons.org/licenses/by-nc-nd/2.5/co/
dc.rights.local.spa.fl_str_mv Abierto (Texto Completo)
dc.rights.accessrights.spa.fl_str_mv info:eu-repo/semantics/openAccess
http://purl.org/coar/access_right/c_abf2
dc.rights.creativecommons.*.fl_str_mv Atribución-NoComercial-SinDerivadas 2.5 Colombia
rights_invalid_str_mv http://creativecommons.org/licenses/by-nc-nd/2.5/co/
Abierto (Texto Completo)
http://purl.org/coar/access_right/c_abf2
Atribución-NoComercial-SinDerivadas 2.5 Colombia
eu_rights_str_mv openAccess
dc.format.mimetype.spa.fl_str_mv application/pdf
dc.coverage.campus.spa.fl_str_mv UNAB Campus Bucaramanga
dc.publisher.grantor.spa.fl_str_mv Universidad Autónoma de Bucaramanga UNAB
dc.publisher.faculty.spa.fl_str_mv Facultad Ingeniería
dc.publisher.program.spa.fl_str_mv Pregrado Ingeniería de Sistemas
institution Universidad Autónoma de Bucaramanga - UNAB
bitstream.url.fl_str_mv https://repository.unab.edu.co/bitstream/20.500.12749/7053/1/2019_Tesis_YharyEstefania_Arias_Trillos.pdf
https://repository.unab.edu.co/bitstream/20.500.12749/7053/2/2019_Licencia_YharyEstefania_Arias_Trillos.pdf
https://repository.unab.edu.co/bitstream/20.500.12749/7053/3/license.txt
https://repository.unab.edu.co/bitstream/20.500.12749/7053/4/2019_Tesis_YharyEstefania_Arias_Trillos.pdf.jpg
https://repository.unab.edu.co/bitstream/20.500.12749/7053/5/2019_Licencia_YharyEstefania_Arias_Trillos.pdf.jpg
bitstream.checksum.fl_str_mv 815ec3b92e577e557b0fa64e12efed4f
6d651538488462f7aa716d7799cf87ee
8a4605be74aa9ea9d79846c1fba20a33
653fc4450edd1bb858f6196ff6a5a77b
e8b0c57249de02005e02c1fabed3b95a
bitstream.checksumAlgorithm.fl_str_mv MD5
MD5
MD5
MD5
MD5
repository.name.fl_str_mv Repositorio Institucional | Universidad Autónoma de Bucaramanga - UNAB
repository.mail.fl_str_mv repositorio@unab.edu.co
_version_ 1814278442523295744
spelling Ortiz Beltrán, Ariel6f2f1af7-c26f-4acf-88f4-0fdf0fba7842-1Arias Trillos, Yhary Estefanía0bfe4e54-c4a9-465d-9e9a-7b06c7cfbf24-1Ortiz Beltrán, Ariel [0001459925]Ortiz Beltrán, Ariel [Ariel-Ortiz-Beltran]Grupo de Investigación Tecnologías de Información - GTIGrupo de Investigaciones Clínicas2020-07-27T19:19:09Z2020-07-27T19:19:09Z2019http://hdl.handle.net/20.500.12749/7053instname:Universidad Autónoma de Bucaramanga - UNABreponame:Repositorio Institucional UNABrepourl:https://repository.unab.edu.coLa detección de cáncer de tiroides es un proceso que en la actualidad se realiza mediante la interpretación manual que realizan radiólogos especialistas, estas se clasifican utilizando una prueba de tamizaje (discriminatoria) conocida como EU- TIRADS 2017 [2], que determina el grado de malignidad del nódulo tiroideo. La escasez de profesionales y la creciente demanda de este tipo de estudios plantea el problema de la automatización a través de algoritmos de aprendizaje de máquina como los basados en Deep Learning y específicamente, las Redes Neuronales Convolucionales, que han sido probadas anteriormente con éxito para la clasificación de otro tipo de imágenes médicas. En un trabajo anterior, con un dataset de 2000 imágenes balanceado entre 4 categorías (TI-RADS2 - TI-RADS5) se logró una medida de precisión (accuracy) cercana del 65% y una pérdida logarítmica (cross-entropy loss) cercana a 0.78. Sin embargo, este artículo plantea el estudio exploratorio para una posible optimización del algoritmo a través de diferentes pruebas medibles en su parametrización. Las variables que serán ajustadas son: El número de capas convolucionales, el tamaño de la máscara de convolución, las funciones de activación, el número de neuronas en la capa densa, el uso de más capas densas para el aprendizaje, el uso de dropouts aleatorios para controlar el sobreajuste (overfitting), entre otros. La medición comparativa se realiza a través de los valores de precisión, pérdida, la matriz de confusión, y el área bajo la curva ROC. Al final del documento se describe la mejor combinación de los parámetros evaluados y las observaciones pertinentes.Resumen 8 Abstract 8 Introducción 9 Planteamiento del problema 9 Pregunta de investigación 10 Objetivos 10 Objetivo General 10 Objetivos específicos 10 Revisión de literatura 11 Tabla de proyectos 11 Tabla de aplicaciones: 14 Tabla de artículos: 16 Estado del arte 17 Proyectos: 17 Aplicaciones 22 Artículos 26 Marco Teórico 30 Manejo de valores nulos 31 Imputación 32 Estandarización 32 Manejo de variables categóricas 32 La multicolinealidad y su impacto 33 Teorema de Fourier 33 Ultrasonido 34 Extracción de características 37 Pruebas en imágenes 38 Gammagrafía 38 Métricas para evaluar un algoritmo de aprendizaje 39 Precisión de Clasificación 39 Pérdida logarítmica 39 Matriz de confusión 39 Algoritmos de predicción 40 Área bajo la curva 41 Aprendizaje de maquina 42 Metodología 44 Tipo de estudio 45 Población/muestra de referencia 45 Muestra elegible 46 Procesamiento de la imagen 46 Análisis de la imagen: 48 Arquitectura de la red neuronal CNN 48 Entrenamiento de la red neuronal 49 Mediciones de desempeño 50 Resultados 51 Resultados de la prueba 52 Conclusiones 53 Cronograma de actividades 54 Actividades 54 Presupuesto 56 Equipo necesario para el desarrollo 56 Costos: 56 Herramientas: 57 Glosario 58 Referencias 62PregradoDetection of thyroid cancer is a process which is currently done through manual interpretation who perform specialist radiologists, these are classified using a screening test (discriminatory) known as EU- TIRADS 2017 [2], which determines the degree of malignancy of the thyroid nodule. The shortage of professionals and the growing Demand for this type of study raises the problem of automation through machine learning algorithms such as those based on Deep Learning and specifically, Networks Convolutionary neurons, which have been tested formerly successfully for the classification of another type of medical images In a previous job, with a 2000 dataset balanced images between 4 categories (TI-RADS2 - TI-RADS5) a measure of accuracy (accuracy) close to 65% was achieved and a logarithmic loss (cross-entropy loss) close to 0.78. Without However, this article raises the exploratory study for a possible algorithm optimization through different tests Measurable in its parameterization. The variables that will be Fitted are: The number of convolutional layers, the size of the convolution mask, the activation functions, the number of neurons in the dense layer, using more dense layers for the learning, the use of random dropouts to control the overfitting, among others. The comparative measurement is performs through the values of precision, loss, the matrix of confusion, and the area under the ROC curve. At the end of the document describes the best combination of the parameters evaluated and the relevant observations.Modalidad Presencialapplication/pdfspahttp://creativecommons.org/licenses/by-nc-nd/2.5/co/Abierto (Texto Completo)info:eu-repo/semantics/openAccesshttp://purl.org/coar/access_right/c_abf2Atribución-NoComercial-SinDerivadas 2.5 ColombiaSistema web de reconocimiento y clasificación de patologías a través de imágenes médicas basado en técnicas de aprendizaje de máquinaWeb recognition and classification system pathologies through medical images based on machine learning techniquesIngeniero de SistemasUniversidad Autónoma de Bucaramanga UNABFacultad IngenieríaPregrado Ingeniería de Sistemasinfo:eu-repo/semantics/bachelorThesisTrabajo de Gradohttp://purl.org/coar/resource_type/c_7a1fhttp://purl.org/redcol/resource_type/TPSystems engineerTechnological innovationsData setDeep learningMachingData increaseNeural networksArtificial intelligenceUltrasoundRadiologyArea under the curveConfusion matrixConcordance studyCancer diagnosisX-raysMedical examinationsDiagnostic serviceEndocrine glandsIngeniería de sistemasInnovaciones tecnológicasInteligencia artificialCáncer diagnósticoRayos xExámenes médicosServicio de diagnósticoGlándulas endocrinasConjunto de datosAprendizaje profundoAumento de datosRedes neuronalesUltrasonidoRadiologíaÁrea bajo la curvaMatriz de confusiónEstudio de concordancia[1] Chapter 1: Supervised Learning and Naive Bayes Classification — Part 2 (Coding). (2017). Retrieved from https://medium.com/machine-learning101/chapter-1supervised-learning-and-naive-bayes-classification-part-2coding-5966f25f1475[2] clinicalkey.es. (2018). Thyroid Imaging. clinicalkey. Retrieved from https://wwwclinicalkey-es.aure.unab.edu.co/service/content/pdf/watermarked/3s2.0B9780323189071000792.pdf?locale=es_ES[3] clinicalkey.es. (2018). Thyroid Neoplasia. Retrieved from https://wwwclinicalkeyes.aure.unab.edu.co/service/content/pdf/watermarked/3s2.0B9780323189071000925.pdf?locale=es_ES[4] ClinicalKey.es. (2018). The Thyroid Gland. Estados Unidos. Retrieved from https://www-clinicalkey- es.aure.unab.edu.co/service/content/pdf/watermarked/3- s2.0B9780323401715000195.pdf?locale=es_ES[5] Department of Computer Science. (1999). Correlation-based Feature Selection for Machine Learning. Hamilton, NewZealand. Retrieved from https://www.lri.fr/~pierres/donn%E9es/save/these/articles/lprqueue/hall99correlationb ased.pdf[6] Department of Computer Science University of Waikato. (1999). Feature Selection for Machine Learning: Comparing a Correlation-based Filter Approach to the Wrapper. New Zealand: waikato. Retrieved from http://www.aaai.org/Papers/FLAIRS/1999/FLAIRS99-042.pdf[7] ESMERILADO - Definición - Significado. (2018). Retrieved from https://diccionario.motorgiga.com/esmerilado[8] Exámenes PET CT y RM: ¿Qué son y cómo usarlos en la medicina diagnóstica?. (2015). Retrieved from http://www.mv.com.br/es/blog/examenes- -pet-ct-y-rm---que-son-y-como-usarlos-en-la-medicina-diagnosticar[9] Medicina nuclear: SPECT y PET en tumores primarios del Sistema Nervioso | NeuroWikia. (2010). Retrieved from http://www.neurowikia.es/content/medicina-nuclear-spect-y-pet-en-tumoresprimariosdel-sn[10] Northside Radiology Associates. (2016). MIBG scintiscan. Atlanta: Editorial team. Retrieved from https://medlineplus.gov/ency/article/003830.htm[11] Positron. (2018). Retrieved from https://en.wikipedia.org/wiki/Positron[12] Pruebas para detectar el cáncer de tiroides. (2016). Retrieved from https://www.cancer.org/es/cancer/cancer-de-tiroides/detecciondiagnosticoclasificacion-por-etapas/como-se-diagnostica.html[13] University of Washington School of Medicine. (2016). Gammagrafía de la tiroides. Seattle: Editorial Director y A.D.A.M. Editorial team. Retrieved from https://medlineplus.gov/spanish/ency/article/003829.htm[14] working paper series. (2000). Correlation - based feature selection ffor discrete and numeric class Machine Learning. New Zealand. Retrieved from https://researchcommons.waikato.ac.nz/bitstream/handle/10289/1024/uow-cswp- 2000-08.pdf?sequence=1&isAllowed=y[15] Yodo-131. (2018). Retrieved from https://es.wikipedia.org/wiki/Yodo-131[16] ¿Qué es una API REST? - Idento. (2018). Retrieved from https://www.idento.es/blog/desarrollo-web/que-es-una-api-rest/[17] Cloud AutoML - Custom Machine Learning Models | AutoML | Google Cloud. (2018). Retrieved from https://cloud.google.com/automl/[18] Convolutional Neural Networks for Visual Recognition. (2018). Retrieved from http://cs231n.github.io/convolutional-networks/[19] general, M. (2018). Salarios de Médico/a general en Colombia | Indeed.com. Retrieved from https://co.indeed.com/salaries/M%C3%A9dico/a-generalSalaries[20] Journal of Physics: Conference Series. (2016). Recent development of feature extraction and classification multispectral/hyperspectral images: a systematic literature review. Retrieved from http://iopscience.iop.org/article/10.1088/17426596/801/1/012045/pdf[21] Judith Marcin, M. (2018). MRI Scans: Definition, uses, and procedure. Retrieved from https://www.medicalnewstoday.com/articles/146309.php[22] Judith Marcin, M. (2018). MRI Scans: Definition, uses, and procedure. Retrieved from https://www.medicalnewstoday.com/articles/146309.php[21] Principal Component Analysis [PCA]. (2017). Retrieved from https://medium.com/100-days-of-algorithms/day-92-pca-bdb66840a8fb[22] ProClass Software - 2018 Reviews, Pricing & Demo. (2018). Retrieved from https://www.softwareadvice.com/registration/proclass-profile/[23] ProClass Software - 2018 Reviews, Pricing & Demo. (2018). Retrieved from https://www.softwareadvice.com/registration/proclass-profile/[24] Purves, D., Augustine, G., Fitzpatrick, D., Katz, L., LaMantia, A., McNamara, J., & Williams, S. (2018). Types of Eye Movements and Their Functions. Retrieved from https://www.ncbi.nlm.nih.gov/books/NBK10991/[25] Robust image hashing using ring partition-PGNMF and local features. (2016). Retrieved from https://www.ncbi.nlm.nih.gov/pmc/articles/PMC5118381/[26] School of Electronics and Control Engineering Chang'an. (2012). Hyperspectral Image Classification Based on Spectral-Spatial Feature Extraction. Xi’an, China. Retrieved from https://ieeexplore- ieeeorg.aure.unab.edu.co/stamp/stamp.jsp?tp=&arnumber=7958808&tag=1[27] sistemas, I. (2018). Salarios de Ingeniero/a en sistemas en Colombia | Indeed.com. Retrieved from https://co.indeed.com/salaries/Ingeniero/a-ensistemas-Salaries[26] What is a data set? - Quora. (2017). Retrieved from https://www.quora.com/What-is-adata-set[27] Brownlee, J. (2016). What is Deep Learning? Retrieved from https://machinelearningmastery.com/what-is-deep-learning[27] Brownlee, J. (2016). What is Deep Learning? Retrieved from https://machinelearningmastery.com/what-is-deep-learning[29] Raicea, R. (2018). Want to know how Deep Learning works? Here’s a quick guide for everyone. Retrieved from https://medium.freecodecamp.org/want-toknow-how-deeplearning-works-heres-a-quick-guide-for-everyone1aedeca88076[30] Tiempo, C. (2018). Sueldo de un profesional con posgrado. Retrieved from https://www.portafolio.co/economia/empleo/un-trabajador-con-posgrado-ganaenpromedio-3-3-millones-mas-que-un-bachiller-512462[31] School of Electronics and Control Engineering Chang'an University (2017). Hyperspectral Image Classification Based on SpectralSpatial Feature Extraction. Xi’an, China. Retrieved from https://ieeexplore-ieeorg.aure.unab.edu.co/stamp/stamp.jsp?tp=&arnumber=7958808[32] School of Electronics and Control Engineering. (2013). Using Nonnegative Matrix Factorization with Projected Gradient for Hyperspectral Images Feature Extraction. Retrieved from https://ieeexplore-ieee- org.aure.unab.edu.co/stamp/stamp.jsp?tp=&arnumber=6566423&ta g=1[33] Dharmsinh Desai University. (2009). Project Classification Using Soft Computing. Nadiad, Gujarat. Retrieved from https://ieeexplore- ieeeorg.aure.unab.edu.co/stamp/stamp.jsp?tp=&arnumber=5376511&ta g=1[34] Dharmsinh Desai University. (2009). Project Classification Using Soft Computing. Nadiad, Gujarat. Retrieved from https://ieeexplore- ieeeorg.aure.unab.edu.co/stamp/stamp.jsp?tp=&arnumber=5376511&ta g=1[35] Department of Computer Science California State Polytechnic University. (2016). Oculomotor Plant Feature Extraction from Human Saccadic Eye Movements. Pomona, USA. Retrieved from https://ieeexplore-ieee- org.aure.unab.edu.co/stamp/stamp.jsp?tp=&arnumber=8424698[36] Bharathidasan College of Arts & Science. (2016). Fusion of Big Data and Neural Networks for Predicting Thyroid. Tamilnadu, India. Retrieved from https://ieeexplore-ieee- org.aure.unab.edu.co/stamp/stamp.jsp?tp=&arnumber=7955223[36] Bharathidasan College of Arts & Science. (2016). Fusion of Big Data and Neural Networks for Predicting Thyroid. Tamilnadu, India. Retrieved from https://ieeexplore-ieee- org.aure.unab.edu.co/stamp/stamp.jsp?tp=&arnumber=7955223[37] Bernard, O., Lalande, A., Zotti, C., & Cervenansky, F. (2018). Deep Learning Techniques for Automatic MRI Cardiac Multi-structures Segmentation and Diagnosis: Is the Problem Solved? Retrieved from https://ieeexplore-ieee- org.aure.unab.edu.co/stamp/stamp.jsp?tp=&arnumber=8360453[38] School of Informatics and Computing, Indiana University. (2015). Temporal Pattern and Association Discovery of Diagnosis Codes using Deep Learning. IN, USA. Retrieved from https://ieeexploreieee- org.aure.unab.edu.co/stamp/stamp.jsp?tp=&arnumber=7349719&ta g=1[39] Amazon Rekognition – Videos e imágenes – AWS. (2018). Retrieved from https://aws.amazon.com/es/rekognition/?hp=tile&soexp=below[40] AMI de aprendizaje profundo de Amazon. (2018). Retrieved from https://aws.amazon.com/es/machine-learning/amis/?hp=tile&soexp=below[41] Amazon Polly. (2018). Retrieved from https://aws.amazon.com/es/polly/?hp=tile&so-exp=below pruébelo, V. (2018). Chatbot | Deep learning | Amazon Lex. Retrieved from https://aws.amazon.com/es/lex/?hp=tile&soexp=below[42] Amazon Translate – Traducción de máquina neural – AWS. (2018). Retrieved from https://aws.amazon.com/es/translate/?hp=tile&soexp=below[43] What is Watson. (2018). Retrieved from https://www.ibm.com/watson/about/index.html ©2018 IEEE. (2018). Delta Univ. for Science and Technology. Gamasa City, Egypt, from https://ieeexplore-ieee- org.aure.unab.edu.co/stamp/stamp.jsp?tp=&arnumber=8358209[44] National Chung Hsing University. (2009). Optimal Grouping by using Genetic Algorithm and Support Vector Machines. Retrieved from https://ieeexplore-ieee- org.aure.unab.edu.co/stamp/stamp.jsp?tp=&arnumber=5420079[45] EEE TRANSACTIONS ON BIOMEDICAL ENGINEERING. (2010). Thyroid Segmentation and Volume Estimation in Ultrasound Images. Retrieved from https://ieeexplore-ieee- org.aure.unab.edu.co/stamp/stamp.jsp tp=&arnumber=5415666&tag=1[46] IEEE. (2017). Tiroid Kanserinde BilgisayarlÕ Tomografi Temelli Yeni Öznitelikler Computerized Tomography Based Novel Features in Thyroid Cancer. Retrieved from https://ieeexplore-ieee- org.aure.unab.edu.co/stamp/stamp.jsp?tp=&arnumber=8238050[47] IEEE. (2009). Computer-Aided Diagnosis of Thyroid Malignancy Using an Artificial Immune System Classification Algorithm. Retrieved from https://ieeexplore-ieee- org.aure.unab.edu.co/stamp/stamp.jsp?tp=&arnumber=4539694[48] AIMBE. (2012). Automated Benign & Malignant Thyroid Lesion Characterization and Classification in 3D Contrast-Enhanced Ultrasound. IEEE. Retrieved from https://ieeexplore-ieee- org.aure.unab.edu.co/stamp/stamp.jsp?tp=&arnumber=6345965[49] The department of thyroid surgery. (2012). Discussion about misdiagnosed reasons and reoperation of thyroid cancer. Changchun city, China. Retrieved from https://ieeexplore- ieeeorg.aure.unab.edu.co/stamp/stamp.jsp?tp=&arnumber=6291383[50] Department of Electronics and Computer Engineering, Gifu University Yanagido. (1992). Neural Network Approach for the ComputerAided Diagnosis of Coronary Artery Diseases in Nuclear Medicine. Retrieved from https://ieeexplore-ieeeorg.aure.unab.edu.co/stamp/stamp.jsp?tp=&arnumber=227168[51] Shandong Provincial Key Laboratory of Computer Networks. (2016). Web Identification Image Recognition Based on Deep Learning. Jinan, China. Retrieved from https://ieeexplore-ieeeorg.aure.unab.edu.co/stamp/stamp.jsp?tp=&arnumber=7726261[52] College of Computer Science and Technology. (2012). Interface Schema Matching with the Machine Learning for Deep Web. Harbin, P.R. China. Retrieved from https://ieeexplore-ieee.org.aure.unab.edu.co/stamp/stamp.jsp?tp=&arnumber=6526056[53] Towards Data Science. (2019). Metrics to Evaluate your Machine Learning Algorithm. [online] Available at: https://towardsdatascience.com/metrics-to-evaluate-your-machinelearning-algorithm-f10ba6e38234 [Accessed 5 Jun. 2019].UNAB Campus BucaramangaORIGINAL2019_Tesis_YharyEstefania_Arias_Trillos.pdf2019_Tesis_YharyEstefania_Arias_Trillos.pdfTesisapplication/pdf1653868https://repository.unab.edu.co/bitstream/20.500.12749/7053/1/2019_Tesis_YharyEstefania_Arias_Trillos.pdf815ec3b92e577e557b0fa64e12efed4fMD51open access2019_Licencia_YharyEstefania_Arias_Trillos.pdf2019_Licencia_YharyEstefania_Arias_Trillos.pdfLicenciaapplication/pdf119824https://repository.unab.edu.co/bitstream/20.500.12749/7053/2/2019_Licencia_YharyEstefania_Arias_Trillos.pdf6d651538488462f7aa716d7799cf87eeMD52metadata only accessLICENSElicense.txtlicense.txttext/plain; charset=utf-81748https://repository.unab.edu.co/bitstream/20.500.12749/7053/3/license.txt8a4605be74aa9ea9d79846c1fba20a33MD53open accessTHUMBNAIL2019_Tesis_YharyEstefania_Arias_Trillos.pdf.jpg2019_Tesis_YharyEstefania_Arias_Trillos.pdf.jpgIM Thumbnailimage/jpeg4881https://repository.unab.edu.co/bitstream/20.500.12749/7053/4/2019_Tesis_YharyEstefania_Arias_Trillos.pdf.jpg653fc4450edd1bb858f6196ff6a5a77bMD54open access2019_Licencia_YharyEstefania_Arias_Trillos.pdf.jpg2019_Licencia_YharyEstefania_Arias_Trillos.pdf.jpgIM Thumbnailimage/jpeg11140https://repository.unab.edu.co/bitstream/20.500.12749/7053/5/2019_Licencia_YharyEstefania_Arias_Trillos.pdf.jpge8b0c57249de02005e02c1fabed3b95aMD55metadata only access20.500.12749/7053oai:repository.unab.edu.co:20.500.12749/70532024-01-19 18:57:20.413open accessRepositorio Institucional | Universidad Autónoma de Bucaramanga - UNABrepositorio@unab.edu.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