Figura 26 - uploaded by Jorge G Pires
Content may be subject to copyright.
Comparação entre modelos baseados "análise do dados" (e.x., segmentação) e "modelagem dos dados" (e.x., equações diferenciais). Fonte: CSDS short-course "Data Science and Biomathematics: an introduction to mathematical modeling applied to biological systems with Matlab" Na figura abaixo, Figura 27, tem-se o uso de modelos matemáticos para estimar parâmetros em modelos matemáticos.
Source publication
O emprego de imagens está cada vez mais presente nos tempos atuais, especialmente nas áreas médicas e de visão computacional; e segmentação de imagens é uma metodologia largamente empregada para se fazer bom uso dessas imagens. Conceitualmente, segmentação de imagens é parte de quase todos esquemas computacionais para reconhecimento de padrão em im...
Citations
... Most of the routine tasks done by medical doctors can be automated, or are on the verge of, such as reading X-ray plates. We have organized a couple here [26]. Another benefit could actually reducing be costs: one benefice of AI models is that intelligence starts to be cheaper, once the model learns, it can be easily shared as API, as an example. ...
Contributions to medicine may come from different areas, and most of these areas are filled with researchers eager to contribute. In this paper, we aim to contribute through the intersection of machine learning and web development. We employed TensorFlow.js, a JavaScript-based library, to model biomedical datasets using neural networks obtained from Kaggle. The principal aim of this study is to present the capabilities of TensorFlow.js and promote its utility in the development of sophisticated machine learning models customized for web-based applications. We modeled three datasets: diabetes detection, surgery complications, and heart failure. While Python and R currently dominate, JavaScript and its derivatives are rapidly gaining ground, offering comparable performance and additional features associated with JavaScript. Kaggle, the public platform from which we downloaded our datasets, provides an extensive collection of biomedical datasets. Therefore, readers can easily test our discussed methods by using the provided codes with minor adjustments on any case of their interest. The results demonstrate an accuracy of 92% for diabetes detection, almost 100% for surgery complications, and 80% for heart failure. The possibilities are vast, and we believe that this is an excellent option for researchers focusing on web applications, particularly in the field of medicine.
... As one example on how this tool can be powerful, we have Pires (2018) reported an example where researchers applied a general-trained based model for teeth segmentation, making it possible to achieve excellent results from a small number of images. We are using the same approach, except they had to build the model from scratch, requiring a couple of PhD students and masters in computer science as so they could make the transfer learning. ...
Introduction: deep learning emerged in 2012 as one of themost important machine learning technologies, reducing image identification error from25% to 5%. This article has two goals: 1) to demonstrate to the general public the ease of building state-of-the-art machine learningmodels without coding expertise; 2) to present a basicmodel adaptable to any biological image identification, such as species identification. Method: We present three test-of-conceptmodels thatshowcase distinct perspectives of the app. Themodels aim at separating images into classes such as genus, species, and subspecies, and the input images can be easily adapted for different cases. We have applied deep learning and transfer learning using TeachableMachine. Results: Our basicmodels demonstrate high accuracy in identifying different speciesbased on images, highlighting the potential for thismethod to be applied in biology. Discussions: the presentedmodels showcase the ease of using machine learning nowadays for image identification. Furthermore, the adaptability of this method to various species and genuses emphasizes its importance in the biological fields, as root for inspiring collaborations with computer science. On our, future collaborations could lead to increasingly accurate and efficientmodels in this arena using well-curated datasets.
... Most of the routine tasks done by medical doctors can be automated, or are on the verge of, such as reading X-ray plates. We have organized a couple here [17]. Another benefit could actually reducing costs: one benefice of AI models is that intelligence starts to be cheap, once the model learns, it can be easily share as API, as an example. ...
Introduction
Contributions to medicine may come from different areas; and most areas are full of researchers wanting to support. Physists may help with theory, such as for nuclear medicine. Engineers with machineries, such as dialysis machine. Mathematicians with models, such as pharmacokinetics. And computer scientists with codes such as bioinformatics.
Method
We have used TensorFlow.js for modeling using neural networks biomedical datasets from Kaggle. We have modeled three datasets: diabetes detection, surgery complications, and heart failure. We have used Angular coded in TypeScript for the implementation of the models. Using TensorFlow.js, we have built Multilayer Perceptrons (MPLs) for modelling our datasets. We have employed the training and the validation curves to make sure the model learnt, and we have used accuracy as a measure of goodness of each model.
Results and discussion
We have built a couple of examples using TensorFlow.js as machine learning platform. Even though python and R are dominant at the moment, JavaScript and derivatives are growing fast, offering basically the same performance, and some extra features associated with JavaScript. Kaggle, the public platform from where we downloaded our datasets, offers a huge amount of datasets for biomedical cases, thus, the reader can easily test what we have discussed, using the same codes, with minor chances, on any case they may be interested in. We were able to find 92% of accuracy for diabetes detection, 100% for surgery complications, and 70% for heart failure. The possibilities are unlimited, and we believe that it is a nice option for researchers aiming at web applications, especially, focused on medicine.
Resumo
Palavras-Chave
CC BY-NC-ND 4 . 0 - This work is licensed under a Creative Commons Attribution-NonCommercial-NoDerivatives 4 . 0 International License .
... Most of the routine tasks done by medical doctors can be automated, or are on the verge of, such as reading X-ray plates. We have organized a couple here [17]. Another benefit could actually reducing costs: one benefice of AI models is that intelligence starts to be cheap, once the model learns, it can be easily share as API, as an example. ...
Introduction: Contributions to medicine may come from different areas; and most areas are full of researchers wanting to support. Physists may help with theory, such as for nuclear medicine. Engineers with machineries, such as dialysis machine. Mathematicians with models, such as pharmacokinetics. And computer scientists with codes such as bioinformatics. Method: We have used TensorFlow.js for modeling using neural networks biomedical datasets from Kaggle. We have modeled three datasets: diabetes detection, surgery complications, and heart failure. We have used Angular coded in TypeScript for the implementation of the models. Using TensorFlow.js, we have built Multilayer Perceptrons (MPLs) for modelling our datasets. We have employed the training and the validation curves to make sure the model learnt, and we have used accuracy as a measure of goodness of each model. Results and discussion: We have built a couple of examples using TensorFlow.js as machine learning platform. Even though python and R are dominant at the moment, JavaScript and derivatives are growing fast, offering basically the same performance, and some extra features associated with JavaScript. Kaggle, the public platform from where we downloaded our datasets, offers a huge amount of datasets for biomedical cases, thus, the reader can easily test what we have discussed, using the same codes, with minor chances, on any case they may be interested in. We were able to find 92% of accuracy for diabetes detection, 100% for surgery complications, and 70% for heart failure. The possibilities are unlimited, and we believe that it is a nice option for researchers aiming at web applications, especially, focused on medicine.
... As one example on how this tool can be powerful, We have [3] reported an example where researchers applied a general-trained based model for teeth segmentation, making it possible to achieve excellent results from a small number of images. We are using the same approach, except they had to build the model from scratch, requiring a couple of PhD students and masters in computer science as so they could make the transfer learning. ...
Introduction: deep learning emerged in 2012 as one of the most important machine learning technologies, reducing image identification error from 25% to 5%. This article has two goals: 1) to demonstrate to the general public the ease of building state-of-the-art machine learning models without coding expertise; 2) to present a basic model adaptable to any biological image identification, such as species identification. Method: We present three test concept models that showcase distinct perspectives of the app. The models aim at separating images into classes such as genus, species, and subspecies, and the input image can be easily adapted for different cases. We have applied deep learning and transfer learning using Teachable Machine. Results: Our basic models demonstrate high accuracy in identifying different species based on images, highlighting the potential for this method to be applied in biology. Discussions: the presented models showcase the ease of using machine learning nowadays for image identification. Furthermore, the adaptability of this method to various species and genuses emphasizes its importance in the biological fields, as root for inspiring collaborations with computer science. Future collaborations could lead to increasingly accurate and efficient models in this arena using well-curated datasets.
Resumo: Introdução: o aprendizagem profundo surgiu em 2012 como uma das tecnologias de aprendizado de máquina mais importantes, reduzindo o erro de identificação de imagem de 25% para 5%. Este artigo tem dois objetivos: 1) demonstrar ao público em geral a facilidade de construir modelos de aprendizado de máquina deúltima geração sem expertise em programação; 2) apresentar um modelo básico adaptável a qualquer identificação de imagem biológica, como identificação de espécies. Método: Apresentamos três modelos conceituais de teste que mostram perspectivas distintas do aplicativo. Os modelos têm como objetivo separar imagens em classes como gênero, espécie e subespécie, e a imagem de entrada pode ser facilmente adaptada para diferentes casos. Aplicamos aprendizado profundo e transferência de aprendizado usando o Teachable Machine. Resultados: Nossos modelos básicos demonstram alta precisão na identificação de diferentes espécies com base em imagens, destacando o potencial deste método para ser aplicado em biologia. Discussões: os modelos apresentados mostram a facilidade de usar o aprendizado de máquina atualmente para identificação de imagens. Além disso, a adaptabilidade deste método a várias espécies e gêneros enfatiza sua importância nos campos biológicos, como base para inspirar colaborações com a ciência da computação. Colaborações futuras podem levar a modelos cada vez mais precisos e eficientes nessaárea usando bancos de dados bem-curados. Palavras-Chave: bioinformática-biologia-serpentes-aprendizado profundo-transferência de conheci-mento-JavaScript
... As one example on how this tool can be powerful, We have [3] reported an example where researchers applied a general-trained based model for teeth segmentation, making it possible to achieve excellent results from a small number of images. We are using the same approach, except they had to build the model from scratch, requiring a couple of PhD students and masters in computer science as so they could make the transfer learning. ...
Introduction
deep learning emerged in 2012 as one of the most important machine learning technologies, reducing image identification error from 25% to 5%. This article has two goals: 1) to demonstrate to the general public the ease of building state-of-the-art machine learning models without coding expertise; 2) to present a basic model adaptable to any biological image identification, such as species identification.
Method
We present three test concept models that showcase distinct perspectives of the app. The models aim at separating images into classes such as genus, species, and subspecies, and the input image can be easily adapted for different cases. We have applied deep learning and transfer learning using Teachable Machine.
Results
Our basic models demonstrate high accuracy in identifying different species based on images, highlighting the potential for this method to be applied in biology.
Discussions
the presented models showcase the ease of using machine learning nowadays for image identification. Furthermore, the adaptability of this method to various species and genuses emphasizes its importance in the biological fields, as root for inspiring collaborations with computer science. Future collaborations could lead to increasingly accurate and efficient models in this arena using well-curated datasets.
... Algumas aplicações e reflexões do uso de processamento de imagens médicas podem ser encontradas em Pires (6) Uma questão interessante do trabalho de Barros (8) , apontado em Pires (6) , ...
... Algumas aplicações e reflexões do uso de processamento de imagens médicas podem ser encontradas em Pires (6) Uma questão interessante do trabalho de Barros (8) , apontado em Pires (6) , ...
Nos últimos anos, tem havido um crescimento, quase exponencial, dos custos com saúde. Como consequência, formas diversas têm sido propostas e implementadas para lidar com as demandas na área de saúde. Nos tempos atuais, estas demandas envolvem tratamentos cada vez mais eficientes, personalizados, acessíveis e de baixo custo. Diante deste cenário, pode-se dizer que as chamadas startups são a grande promessa para lidar com iniciativas, como implementação de novas ideias e produtos. As startups trazem novas formas de pensar e produzir, com um grau de liberdade não presente em grandes empresas. Pretende-se discutir espaços para startups na área de saúde, em diálogo com literaturas disponíveis na internet. O objetivo deste trabalho é evidenciar como as startups podem contribuir para a medicina personalizada alcançar seus objetivos, healthcare, em geral, superando o dilema da ‘personalização vs. custo’. A motivação é que apesar do desenvolvido das universidades, muitos países, sendo o Brasil um desses, não possuem um fluxo contínuo e corriqueiro de transformação de ciência em bens para a sociedade. Conclui-se que as startups se apresentam como a melhor opção para resolver/lidar com vários problemas/dilemas que surgiram nas áreas médicas, nas últimas décadas, como aumento dos custos e demanda cada vez maior por tratamentos especializados/personalidades.
Le post-doctorat est une activité spécialisée en recherche universitaire, réalisée par des professionnels possédant déjà un doctorat et représentant une opportunité d’acquérir des connaissances solides dans un domaine en particulier. Au Brésil, l’une des opportunités postdoctorales les plus connues est proposée par la Coordination et l’Amélioration du Personnel de Niveau supérieur ou du Personnel Enseignant, appelée “Programa Nacional de Pós Doutorado” (PNPD/CAPES). Cette formation a été créée dans le cadre du ‘Programme Régional de Post-Graduation en Développement et Environnement’ de l’Université Fédérale de Paraíba (PRODEMA/UFPB, Brésil) qui assure l’encadrement et le suivi des rapports de masters, thèses doctorales, activités pédagogiques auxiliaires des cours de masters, en complément des activités développées avec des entreprises et des associations. Les connaissances acquises au cours de ce programme assurent une meilleure qualification professionnelle, permettant ainsi de développer des postes pertinents, avec des tâches et des responsabilités inhérentes à un professionnel avec un doctorat.
