{"title":"在智能手机中使用卷积神经网络识别小数据集的口腔疾病","authors":"Jormany Quintero-Rojas, Jesús David González","doi":"10.19053/01211129.V30.N55.2021.11846","DOIUrl":null,"url":null,"abstract":"espanolEl reconocimiento y procesamiento de imagenes es una herramienta adecuada en los sistemas que usan metodos de aprendizaje automatico. La adicion de telefonos inteligentes como herramientas complementarias en el area de la salud para el diagnostico es un hecho hoy en dia por las ventajas que presentan. Siguiendo la tendencia de proporcionar herramientas para el diagnostico, esta investigacion tuvo como objetivo desarrollar una aplicacion movil prototipo para la identificacion de lesiones bucales, incluyendo lesiones potencialmente malignas, basado en redes neuronales convolucionales, como la deteccion temprana de indicios de posibles tipos de cancer en la cavidad bucal. Se desarrollo una aplicacion movil para el sistema operativo Android que implemento la libreria de TensorFlow y el modelo de redes neuronales convolucionales Mobilenet V2. El entrenamiento del modelo se realizo por transferencia de aprendizaje con una base de datos de 500 imagenes distribuidas en cinco clases para el reconocimiento (Leucoplasia, Herpes Simple Virus Tipo 1, Estomatitis aftosa, Estomatitis nicotinica y Sin lesion). Se utilizo el 80% de las imagenes para el entrenamiento y el 20% para la validacion. Se obtuvo que la aplicacion presento al menos 80% de exactitud en el reconocimiento de cuatro clases. Se usaron las metricas de f1-valor y area bajo la curva para evaluar el desempeno. La aplicacion movil desarrollada presento un comportamiento aceptable con metricas mayores al 75% para el reconocimiento de tres lesiones, por otro lado, arrojo un desempeno desfavorable menor al 70% para identificar los casos de estomatitis nicotinica con el conjunto de datos elegido. EnglishImage recognition and processing is a suitable tool in systems using machine learning methods. The addition of smartphones as complementary tools in the health area for diagnosis is a fact nowadays due to the advantages they present. Following the trend of providing tools for diagnosis, this research aimed to develop a prototype mobile application for the identification of oral lesions, including potentially malignant lesions, based on convolutional neural networks, as early detection of indications of possible types of cancer in the oral cavity. A mobile application was developed for the Android operating system that implemented the TensorFlow library and the Mobilenet V2 convolutional neural network model. The training of the model was performed by transfer learning with a database of 500 images distributed in five classes for recognition (Leukoplakia, Herpes Simplex Virus Type 1, Aphthous stomatitis, Nicotinic stomatitis, and No lesion). The 80% of the images were used for training and 20% for validation. It was obtained that the application presented at least 80% precision in the recognition of four class. The f1-score and area under curve metrics were used to evaluate performance. The developed mobile application presented an acceptable performance with metrics higher than 75% for the recognition of three lesions, on the other hand, it yielded an unfavorable performance lower than 70% for identifying nicotinic stomatitis cases with the chosen dataset.","PeriodicalId":21428,"journal":{"name":"Revista Facultad De Ingenieria-universidad De Antioquia","volume":null,"pages":null},"PeriodicalIF":0.9000,"publicationDate":"2021-02-07","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":"1","resultStr":"{\"title\":\"Use of Convolutional Neural Networks in Smartphones for the Identification of Oral Diseases Using a Small Dataset\",\"authors\":\"Jormany Quintero-Rojas, Jesús David González\",\"doi\":\"10.19053/01211129.V30.N55.2021.11846\",\"DOIUrl\":null,\"url\":null,\"abstract\":\"espanolEl reconocimiento y procesamiento de imagenes es una herramienta adecuada en los sistemas que usan metodos de aprendizaje automatico. La adicion de telefonos inteligentes como herramientas complementarias en el area de la salud para el diagnostico es un hecho hoy en dia por las ventajas que presentan. Siguiendo la tendencia de proporcionar herramientas para el diagnostico, esta investigacion tuvo como objetivo desarrollar una aplicacion movil prototipo para la identificacion de lesiones bucales, incluyendo lesiones potencialmente malignas, basado en redes neuronales convolucionales, como la deteccion temprana de indicios de posibles tipos de cancer en la cavidad bucal. Se desarrollo una aplicacion movil para el sistema operativo Android que implemento la libreria de TensorFlow y el modelo de redes neuronales convolucionales Mobilenet V2. El entrenamiento del modelo se realizo por transferencia de aprendizaje con una base de datos de 500 imagenes distribuidas en cinco clases para el reconocimiento (Leucoplasia, Herpes Simple Virus Tipo 1, Estomatitis aftosa, Estomatitis nicotinica y Sin lesion). Se utilizo el 80% de las imagenes para el entrenamiento y el 20% para la validacion. Se obtuvo que la aplicacion presento al menos 80% de exactitud en el reconocimiento de cuatro clases. Se usaron las metricas de f1-valor y area bajo la curva para evaluar el desempeno. La aplicacion movil desarrollada presento un comportamiento aceptable con metricas mayores al 75% para el reconocimiento de tres lesiones, por otro lado, arrojo un desempeno desfavorable menor al 70% para identificar los casos de estomatitis nicotinica con el conjunto de datos elegido. EnglishImage recognition and processing is a suitable tool in systems using machine learning methods. The addition of smartphones as complementary tools in the health area for diagnosis is a fact nowadays due to the advantages they present. Following the trend of providing tools for diagnosis, this research aimed to develop a prototype mobile application for the identification of oral lesions, including potentially malignant lesions, based on convolutional neural networks, as early detection of indications of possible types of cancer in the oral cavity. A mobile application was developed for the Android operating system that implemented the TensorFlow library and the Mobilenet V2 convolutional neural network model. The training of the model was performed by transfer learning with a database of 500 images distributed in five classes for recognition (Leukoplakia, Herpes Simplex Virus Type 1, Aphthous stomatitis, Nicotinic stomatitis, and No lesion). The 80% of the images were used for training and 20% for validation. It was obtained that the application presented at least 80% precision in the recognition of four class. The f1-score and area under curve metrics were used to evaluate performance. The developed mobile application presented an acceptable performance with metrics higher than 75% for the recognition of three lesions, on the other hand, it yielded an unfavorable performance lower than 70% for identifying nicotinic stomatitis cases with the chosen dataset.\",\"PeriodicalId\":21428,\"journal\":{\"name\":\"Revista Facultad De Ingenieria-universidad De Antioquia\",\"volume\":null,\"pages\":null},\"PeriodicalIF\":0.9000,\"publicationDate\":\"2021-02-07\",\"publicationTypes\":\"Journal Article\",\"fieldsOfStudy\":null,\"isOpenAccess\":false,\"openAccessPdf\":\"\",\"citationCount\":\"1\",\"resultStr\":null,\"platform\":\"Semanticscholar\",\"paperid\":null,\"PeriodicalName\":\"Revista Facultad De Ingenieria-universidad De Antioquia\",\"FirstCategoryId\":\"1085\",\"ListUrlMain\":\"https://doi.org/10.19053/01211129.V30.N55.2021.11846\",\"RegionNum\":0,\"RegionCategory\":null,\"ArticlePicture\":[],\"TitleCN\":null,\"AbstractTextCN\":null,\"PMCID\":null,\"EPubDate\":\"\",\"PubModel\":\"\",\"JCR\":\"Q3\",\"JCRName\":\"ENGINEERING, MULTIDISCIPLINARY\",\"Score\":null,\"Total\":0}","platform":"Semanticscholar","paperid":null,"PeriodicalName":"Revista Facultad De Ingenieria-universidad De Antioquia","FirstCategoryId":"1085","ListUrlMain":"https://doi.org/10.19053/01211129.V30.N55.2021.11846","RegionNum":0,"RegionCategory":null,"ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":null,"EPubDate":"","PubModel":"","JCR":"Q3","JCRName":"ENGINEERING, MULTIDISCIPLINARY","Score":null,"Total":0}
引用次数: 1
摘要
在使用机器学习方法的系统中,图像识别和处理是一种合适的工具。由于智能手机的优势,智能手机作为诊断健康领域的辅助工具的增加是当今的事实。按照趋势为diagnostico提供工具,这种调查有其目的是开发一个《手机原型在内的口腔损伤识别潜在的恶性伤害,基于神经网络等convolucionales成为迹象可能口腔癌症类型。为Android操作系统开发了一个移动应用程序,实现了TensorFlow库和卷积神经网络模型Mobilenet V2。该模型的训练是通过学习转移进行的,数据库中有500张图像,分布在5个识别类(白斑、1型单纯疱疹病毒、口腔炎、烟碱和无病变)。80%的图像用于培训,20%用于验证。结果表明,该应用程序在四类识别中准确率至少为80%。采用f1值和曲线下面积指标来评价性能。所开发的移动应用程序在识别三种病变时表现良好,指标大于75%,另一方面,在识别所选数据集的烟碱口炎病例时表现不佳,低于70%。图像识别和处理是使用机器学习方法的系统中的一个合适的工具。由于智能手机带来的优势,智能手机作为健康领域诊断的辅助工具的加入现在是一个事实。根据提供诊断工具的趋势,本研究旨在开发一种原型移动应用程序,用于识别口腔病变,包括潜在的恶性病变,基于卷曲神经网络,作为早期检测口腔可能类型癌症的迹象。Android A mobile application was发达for the operating system that实行the TensorFlow library and the Mobilenet V2 convolutional neural network model。该模型的训练是通过转移学习进行的,该数据库包含500张图像,分布在5个识别类(白斑、单纯疱疹病毒1型、口腔炎、尼古丁口腔炎和无病变)。= =地理= =根据美国人口普查局的数据,该镇总面积为,其中土地和(1.2%)水。据了解,该申请在四类识别方面的准确率至少为80%。= =地理= =根据美国人口普查局的数据,这个县的总面积,其中土地和(1.1%)水。所开发的移动应用程序在识别三种损伤方面表现良好,指标高于75%,而在使用选定数据集识别尼古丁口炎病例方面表现不佳,低于70%。
Use of Convolutional Neural Networks in Smartphones for the Identification of Oral Diseases Using a Small Dataset
espanolEl reconocimiento y procesamiento de imagenes es una herramienta adecuada en los sistemas que usan metodos de aprendizaje automatico. La adicion de telefonos inteligentes como herramientas complementarias en el area de la salud para el diagnostico es un hecho hoy en dia por las ventajas que presentan. Siguiendo la tendencia de proporcionar herramientas para el diagnostico, esta investigacion tuvo como objetivo desarrollar una aplicacion movil prototipo para la identificacion de lesiones bucales, incluyendo lesiones potencialmente malignas, basado en redes neuronales convolucionales, como la deteccion temprana de indicios de posibles tipos de cancer en la cavidad bucal. Se desarrollo una aplicacion movil para el sistema operativo Android que implemento la libreria de TensorFlow y el modelo de redes neuronales convolucionales Mobilenet V2. El entrenamiento del modelo se realizo por transferencia de aprendizaje con una base de datos de 500 imagenes distribuidas en cinco clases para el reconocimiento (Leucoplasia, Herpes Simple Virus Tipo 1, Estomatitis aftosa, Estomatitis nicotinica y Sin lesion). Se utilizo el 80% de las imagenes para el entrenamiento y el 20% para la validacion. Se obtuvo que la aplicacion presento al menos 80% de exactitud en el reconocimiento de cuatro clases. Se usaron las metricas de f1-valor y area bajo la curva para evaluar el desempeno. La aplicacion movil desarrollada presento un comportamiento aceptable con metricas mayores al 75% para el reconocimiento de tres lesiones, por otro lado, arrojo un desempeno desfavorable menor al 70% para identificar los casos de estomatitis nicotinica con el conjunto de datos elegido. EnglishImage recognition and processing is a suitable tool in systems using machine learning methods. The addition of smartphones as complementary tools in the health area for diagnosis is a fact nowadays due to the advantages they present. Following the trend of providing tools for diagnosis, this research aimed to develop a prototype mobile application for the identification of oral lesions, including potentially malignant lesions, based on convolutional neural networks, as early detection of indications of possible types of cancer in the oral cavity. A mobile application was developed for the Android operating system that implemented the TensorFlow library and the Mobilenet V2 convolutional neural network model. The training of the model was performed by transfer learning with a database of 500 images distributed in five classes for recognition (Leukoplakia, Herpes Simplex Virus Type 1, Aphthous stomatitis, Nicotinic stomatitis, and No lesion). The 80% of the images were used for training and 20% for validation. It was obtained that the application presented at least 80% precision in the recognition of four class. The f1-score and area under curve metrics were used to evaluate performance. The developed mobile application presented an acceptable performance with metrics higher than 75% for the recognition of three lesions, on the other hand, it yielded an unfavorable performance lower than 70% for identifying nicotinic stomatitis cases with the chosen dataset.
期刊介绍:
Revista Facultad de Ingenieria started in 1984 and is a publication of the School of Engineering at the University of Antioquia.
The main objective of the journal is to promote and stimulate the publishing of national and international scientific research results. The journal publishes original articles, resulting from scientific research, experimental and or simulation studies in engineering sciences, technology, and similar disciplines (Electronics, Telecommunications, Bioengineering, Biotechnology, Electrical, Computer Science, Mechanical, Chemical, Environmental, Materials, Sanitary, Civil and Industrial Engineering).
In exceptional cases, the journal will publish insightful articles related to current important subjects, or revision articles representing a significant contribution to the contextualization of the state of the art in a known relevant topic. Case reports will only be published when those cases are related to studies in which the validity of a methodology is being proven for the first time, or when a significant contribution to the knowledge of an unexplored system can be proven.
All published articles have undergone a peer review process, carried out by experts recognized for their knowledge and contributions to the relevant field.
To adapt the Journal to international standards and to promote the visibility of the published articles; and therefore, to have a greater impact in the global academic community, after November 1st 2013, the journal will accept only manuscripts written in English for reviewing and publication.
Revista Facultad de Ingeniería –redin is entirely financed by University of Antioquia
Since 2015, every article accepted for publication in the journal is assigned a DOI number.