Laboratory of Image Data Science - LIDS

10/06/2022

O Laboratory of Image Data Science (LIDS) do Instituto de Computação (IC) da Universidade Estadual de Campinas (UNICAMP), coordenado pelos Profs. Alexandre Falcão e Jancarlo Gomes, depositou sua 15ª.

Esta concessão de invenção (BR 102020023751-9 A2) apresenta como depositantes a UNICAMP e ImmunoCamp Ciência e Tecnologia (empresa filha da UNICAMP) e tem como seus inventores o Prof. Alexandre Falcão, o Prof. Jancarlo Gomes, e os pesquiadores Celso Suzuki e Fábio de Freitas.

O produto resultante desta invenção tem o nome de Diagnóstico Automatizado de Parasitos Intestinais (DAPI) e foi financiado pela FAPESP, UNICAMP e ImmunoCamp com materiais permanente e de
consumo, assim como bolsas para alunos. O DAPI acaba de receber uma premiação de propriedade intelectual licenciada da Agência Inova da UNICAMP e contribuirá para a sociedade sobretudo por todo seu ineditismo científico e tecnológico.

Efetivamente, este dispositivo de diagnóstico passa por um processo final de industrialização e será implementado nos principais laboratórios públicos e privados do país, o que deverá gerar pagamentos de royalties por um período de 15 anos a UNICAMP (IC) e FAPESP.

02/07/2021

03/12/2020

Parabéns, Prof. Alexandre Falcão!

Pesquisa publicada pelo Journal Plos Biology em 16 de outubro de 2020 traz um ranking mundial dos cientistas mais influentes.

Esse ranking elenca os 100.000 top cientistas do mundo, segundo os bancos de dados utilizados, até o ano de 2019.

A pesquisa foi conduzida por uma equipe da Universidade de Stanford (EUA), liderada por John Ioannidis e intitulada:

"Updated science-wide author databases of standardized citation indicators" — https://doi.org/10.1371/journal.pbio.3000918

O Instituto de Computação está muito feliz por ter quatro docentes na lista dos mais influentes do mundo. A Unicamp teve, ao todo, 74 pesquisadores na lista, sendo a segunda instituição do Brasil com mais nomes.

Parabéns aos professores, seus alunos e colaboradores por tornarem isso possível:

Prof. Alexandre Falcão
Prof. Anderson Rocha
Prof. Leandro Villas
Prof. Luiz Fernando Bittencourt

07/07/2020

Texto por Ana Paula Palazi Foto por Pedro Amatuzzi Imagine um mundo onde quase metade da população esteja infectada por organismos que invadem o corpo do hospedeiro, se multiplicam, debilitam a pessoa e podem até levá-la à morte. Os parasitos intestinais são um problema de saúde pública e um...

09/12/2019

Let us talk a bit about visual saliency.

Visual saliency can be defined as the property that makes objects stand out from their surroundings, and as a consequence, capture our attention. In image analysis, saliency maps can be used to indicate regions of interest and improve tasks such as object detection, compression, and segmentation. Saliency maps can be expressed by imagens where the value of each pixel represents it's visual importance (see the post images as an example).

Several methods have been proposed in the literature to define visual saliency and their results are consistently satisfactory, especially on natural images (photographs). Most methods make observations about our behavior to estimate objects of interest: Usually, salient objects are focussed, centered and composed of contrasting colors when compared to the other image objects. However, these assumptions are not always extensible for biomedical images: If we are looking for intestinal parasite eggs in f***l exam images, we cannot assume the objects to be centered and focussed. In this regard, LIDS has been working on methods capable of estimating saliency in different image domains.

Despite our algorithm still being in development, early experiments show promising results for parasite-egg detection. Take a look at this post's images: We have placed side-by-side the original image and its respective saliency map (grayscale images where brighter regions are more salient). The algorithm details will be available in a publication after its final version.

Stay tuned for our news, we will keep you informed of our progress. :)

03/12/2019

Animal care is also a matter of Public Health! The gastrointestinal parasites of dogs and cats are cosmopolitan and have great potential to cause disease in humans. These etiological agents are routinely diagnosed using parasitological stool examination due to their low cost, simple procedure and direct evidence. However, reports in the scientific literature indicate that the techniques currently used to perform this exam have low or moderate diagnostic sensitivity. Therefore, thinking on promoting innovation in the parasitological examination of f***s in Veterinary Medicine, one of the research lines of LIDS aims to develop a new parasitological technique of high diagnostic efficacy in parallel to the development of computational programs for analysis of microscope slide originated images and so in the near future enable the Automated Diagnosis of Dogs and Cats Gastrointestinal Parasites.

25/11/2019

Machines can be trained to classify images and then identify tumors in tomography, compositions in rocks or pathologies in optical microscope analysis. This area of artificial intelligence is known as machine learning and has been gaining new applications in recent years. The training of the machine is done through the repetition of images used as examples of a particular context or situation and the proper preparation of this material requires an effort of experts from the most diverse areas.

"The human is the one who coordinates. Without expert control over the training process, the machine can learn to make decisions based on the characteristics of the image that are not related to the target problem. This generates a bad or restricted result to that database in which the machine was trained. When the database changes, the error increases significantly, making the analysis of the machine reliable ", said Alexandre Xavier Falcão, from the Institute of Computing at University of Campinas (Unicamp), in a lecture presented last Thursday (21/11 ) at Fapesp week France.

Falcão has combined computer science with different areas of knowledge from projects in machine learning, developed with the support of Fapesp, research line that investigates human-machine interaction in decision making.

More information about the project developed in LIDS can be found on the link in Portuguese: http://agencia.fapesp.br/avanco-em-aprendizado-de-maquina-possibilita-novas-tecnologias-baseadas-na-analise-de-imagens/31995/

https://www.youtube.com/watch?v=Y9uUW2tDAUg&feature=emb_title

Technical Session 1: Machine Learning and Data Science Interactive Machine Learning for Image Data Science We aim at training machines to extract information...

19/11/2019

In LIDS, the approach applied research and multidisciplinary is used, in the view that the results achieved of the studies carried out in this academic environment are focused on the desires of society. In this laboratory we seek the interaction of experts from several areas of knowledge (computing, Biology, medicine, chemistry, engineering, etc...) in order to solve problems and difficulties of certain performance areas of great professional and social importance. Multidisciplinary knowledge is generated from the interaction and sum of knowledge between experts from different areas with a same purpose. Multidisciplinary interaction is paramount importance for the confronting of paradigms of the scientific field in the search for innovations and technology, which we call as "State of art" in science.

The results generated by research developed in the environment of LIDS, usually published in scientific articles also exposed on this social page, show mostly this scientific research profile bias.

Follow our networks and follow our page: http://lids.ic.unicamp.br/

Thank you!!

11/11/2019

Our post today is about high dimensional data visualization tools.

When creating machine learning models, it is very important to understand how data is distributed in space. Usually these spaces have a large number of dimensions, which makes it virtually impossible to see. Our brain works well with up to 3 dimensions, it gets much harder with more than that. To visualize this data, we need techniques that can map from a larger dimensionality space to a smaller dimensionality space.

Many machine learning approaches (such as deep learning) work by changing the data feature spaces. We can use visualization tools to try to validate hypotheses about these feature spaces. For example, imagine we have a convolutional neural network to classify whether an image contains a cat or not. If we visualize the resources produced by the convolutional layers, we expect to see images of the same class next to each other and images of different classes far apart. If it does not, we may have a problem with our model.

The most commonly used tools are PCA, tSNE and UMAP, but there are many others. Here are some links about these techniques:
https://towardsdatascience.com/the-art-of-effective-visualization-of-multi-dimensional-data-6c7202990c57

https://medium.com/.mail/pca-principal-component-analysis-729068e28ec8

https://medium.com/.andre/an-introduction-to-t-sne-with-python-example-47e6ae7dc58f

https://towardsdatascience.com/how-exactly-umap-works-13e3040e1668

Strategies for Effective Data Visualization

04/11/2019

Have you heard about Cryptosporidiosis? Not yet?

So don't leave our page without knowing a little more about this important zoonosis.

Zoonoses are diseases of non-human animals that may be transmitted to HUMANS or may be transmitted from humans to non-human animals.. Cryptosporidium spp. , is one of the zoonotic agents that receives special attention in public health because it causes severe diarrhea, dehydration and is among the leading causes of death in children under 5 years old worldwide, according to World Health Organization (WHO).

The recommended diagnosis in the context of Public Health for this agent is the making of permanent slides using Kinyoun's stain, modified Ziehl-Neelsen stain and related methods for visualization of the elimination structures of this protozoan, which we call oocysts.

Currently, LIDS has a research line funded by the São Paulo Research Foundation (FAPESP) that aims to validate in the laboratory the Automated Diagnosis of Intestinal Parasites (DAPI) system for the detection of Cryptosporidium spp. from the adaptation experiments of the TF-Test Coccidia technique, preparing nonpermanent slides and therefore less laborious to be performed . Then, the preparation of an image bank to train the pattern recognition module of DAPI System.
This will bring innovation in microscopic diagnosis by improving techniques that have undergone few changes in the last 30 years.

Want to know more? Visit the link: https://lids.ic.unicamp.br/researches for these and other LIDS research lines!

Já ouviu falar sobre Criptosporidiose? Ainda não?

Então não saia da nossa página sem saber um pouco mais sobre essa importante zoonose.
Zoonoses são doenças que podem ser transmitidas dos humanos aos animais ou dos animais para os humanos. Cryptosporidium spp. , é um dos agentes zoonóticos que recebe especial atenção na Saúde Pública por causar intensa diarreia, desidratação e figurar dentre as principais causas de mortes em crianças abaixo dos 5 anos de idade em todo mundo segundo a Organização Mundial de Saúde (OMS).

O diagnóstico recomendado em contexto de Saúde Pública para este agente é a confecção de lâminas permanentes utilizando os métodos de Kinyoun, Ziehl-Neelsen modificado e métodos afins para a visualização das estruturas de eliminação deste protozoário, os quais chamamos de oocistos.

Atualmente, o LIDS possui uma linha de pesquisa financiada pela Fundação de Amparo à Pesquisa do Estado de São Paulo (FAPESP) que tem como objetivo validar em laboratório o sistema de Diagnóstico Automatizado de Parasitos Intestinais (DAPI) para a detecção de Cryptosporidium spp. a partir de experimentos de adequação da técnica TF- Test Coccidia, preparando lâminas não permanentes e portanto menos trabalhosas de serem realizadas e da preparação de um banco de imagens para treinar o módulo de reconhecimento de padrões do Sistema DAPI. Isso trará inovação no diagnóstico microscópico, aprimorando técnicas que passaram por poucas alterações nos últimos 30 anos.

Quer saber mais? Acesse o link:
https://lids.ic.unicamp.br/researches para essas e outras linhas de pesquisa do LIDS!

29/10/2019

Good afternoon everyone!
We bring in today's post great news about UNICAMP's children companies, those companies that were founded inside the university.
According to data released by the Innovation Agency Inova UNICAMP, the annual turnover of this type of company increased 64.6% only in the previous year. That means an incredible evolution of R$ 3.1 billion.
But what does this mean for the community? One word: Jobs!
That's right! According to data from the same source, university affiliates are responsible for generating over 31,000 direct jobs!
More than 80% of those companies are located in the state of São Paulo (especially in the region of Campinas). Which means that the local population can only benefit from
the development and creation of new children companies.
For more information, see the full article:

https://www.unicamp.br/unicamp/noticias/2019/10/24/faturamento-anual-das-empresas-filhas-da-unicamp-chega-r-79-bilhoes

21/10/2019

Mansonic schistosomiasis, also known as “barriga d’água” or “xistosa” in Brazil, is a disease caused by the parasitic trematode Schistosoma mansoni. The diagnosis is routinely performed by parasitological examination of f***s by conventional quantitative techniques that allow the detection of eggs. The diagnostic sensitivity of these techniques may vary from low to moderate according to several studies in the scientific literature. In this context, a multidisciplinary research is underway in LIDS aiming to develop, evaluate and validate a new parasitological technique called TF-Test Quantified for the quantitative detection of S. mansoni eggs with high sensitivity and specificity in parallel to the development of computational algorithms for automated detection of eggs. This interaction between studies in the fields of parasitology and computer science will allow the automated quantitative diagnosis of S. mansoni eggs. More information on the link: https://onlinelibrary.wiley.com/doi/full/10.1111/tmi.13215

A esquistossomíase mansônica, também conhecida como “barriga d’água” ou “xistosa”, é uma doença causada pelo trematódeo parasito Schistosoma mansoni. O diagnóstico é feito principalmente pelo exame parasitológico de fezes por meio de técnicas convencionais quantitativas que permitem a detecção dos ovos. A sensibilidade diagnóstica dessas técnicas pode variar de baixa a moderada segundo diversos estudos registrados na literatura científica. Diante desse contexto está em andamento no LIDS uma pesquisa multidisciplinar cujo objetivo é desenvolver, avaliar e validar uma nova técnica parasitológica denominada TF-Test Quantified para a detecção quantitativa de ovos de S. mansoni com alta sensibilidade e especificidade em paralelo ao desenvolvimento de algoritmos computacionais para a detecção automatizada dos ovos. Essa interação entre os estudos das áreas da parasitologia e das ciências da computação permitirá de maneira inédita o diagnóstico automatizado quantitativo de ovos de S. mansoni. Mais informações no link: https://onlinelibrary.wiley.com/doi/full/10.1111/tmi.13215