Powered by OpenAIRE graph
Found an issue? Give us feedback
addClaim

Classificação de termogramas do pé diabético utilizando Machine Learning

Authors: Silva, Rita Maria Veiga Almeida e;

Classificação de termogramas do pé diabético utilizando Machine Learning

Abstract

Os pacientes diabéticos são portadores de várias patologias relacionadas com a doença em causa, no entanto uma das principais complicações é o pé diabético. Este está associado ao aumento da temperatura na região plantar que pode implicar a presença ou aparecimento de úlceras devido à presença de neuropatia ou da doença arterial periférica. Uma deteção precoce desta patologia é essencial para diminuir a probabilidade de complicações provocadas por úlceras que podem levar à amputação do pé. Um dos métodos de auxílio mais eficiente e eficaz para detetar problemas relacionados ao pé diabético é a termografia por infravermelho. Sendo que esta permite a profissionais de saúde analisar as distribuições das temperaturas do pé auxiliando assim na perceção de características e padrões de temperaturas intrínsecos ao pé diabético. Técnicas de Inteligência Artificial têm vindo a provarem-se como competentes na área da visão por computador aplicada ao campo da medicina. Estas técnicas são utilizadas no diagnóstico assistido por computador que ajuda a prevenir o erro humano nas decisões tomadas por especialistas. Assim, um sistema de classificação de termografias por infravermelho do pé diabético que utilize técnicas de Inteligência Artificial torna-se uma mais-valia no auxílio ao diagnóstico do mesmo. O presente trabalho explorou três arquiteturas de redes neuronais atualmente usadas para classificação de imagens, a VGG16, VGG19 e a DenseNet. Na aplicação dos modelos utilizados foi utilizada data augmentation e transfer learning de modo colmatar problemas relacionados com a utilização de um dataset pequeno. Posteriormente, foram validados e avaliados os modelos implementados para fazer a classificação binária das termografias por infravermelho do pé, com ou sem risco de ulceração. Os resultados obtidos utilizando estes três modelos são promissores, mas a VGG19 destaca-se com a melhor média de Accurracy em cinco folds definidos na validação do modelo com auxílio à k-fold cross validation. Assim, é demonstrado que a metodologia proposta é eficaz na classificação binária que realiza.

Diabetic patients are carriers of several pathologies related to the disease in question, however one of the main complications is the diabetic foot. This is associated with an increase in temperature in the plantar region, which may imply the presence or appearance of ulcers due to the presence of neuropathy or peripheral arterial disease. An early detection of this pathology is essential to reduce the likelihood of complications caused by ulcers that can lead to amputation of the foot. One of the most efficient and effective support methods to detect problems related to diabetic foot is infrared thermography. This allows health professionals to analyze the distribution of foot temperatures, thus helping to perceive characteristics and temperature patterns intrinsic to the diabetic foot. Artificial Intelligence techniques have been proving to be competent in the field of computer vision applied to medicine. These techniques are used in computer-assisted diagnosis that helps prevent human error in decisions made by experts. Thus, an infrared thermography classification system of the diabetic foot that uses Artificial Intelligence techniques becomes an asset in diagnosis aid. The present work explored three neural network architectures currently used for image classification, VGG16, VGG19 and DenseNet. In the application of the models used, data augmentation and transfer learning were used to solve problems related to the use of a small dataset. Afterwards, the models implemented to perform the binary classification of foot infrared thermographs, with or without risk of ulceration, were validated and evaluated. The results obtained using these three models are promising, but VGG19 stands out with the best average of Accuracy in five defined folds in the model validation with the aid of k-fold cross validation. Thus, it is demonstrated that the proposed methodology is effective in the binary classification it performs.

Dissertação de Mestrado em Engenharia Informática

Country
Portugal
Keywords

Pé diabético, Termografia por infravermelho

  • BIP!
    Impact byBIP!
    selected citations
    These citations are derived from selected sources.
    This is an alternative to the "Influence" indicator, which also reflects the overall/total impact of an article in the research community at large, based on the underlying citation network (diachronically).
    0
    popularity
    This indicator reflects the "current" impact/attention (the "hype") of an article in the research community at large, based on the underlying citation network.
    Average
    influence
    This indicator reflects the overall/total impact of an article in the research community at large, based on the underlying citation network (diachronically).
    Average
    impulse
    This indicator reflects the initial momentum of an article directly after its publication, based on the underlying citation network.
    Average
Powered by OpenAIRE graph
Found an issue? Give us feedback
selected citations
These citations are derived from selected sources.
This is an alternative to the "Influence" indicator, which also reflects the overall/total impact of an article in the research community at large, based on the underlying citation network (diachronically).
BIP!Citations provided by BIP!
popularity
This indicator reflects the "current" impact/attention (the "hype") of an article in the research community at large, based on the underlying citation network.
BIP!Popularity provided by BIP!
influence
This indicator reflects the overall/total impact of an article in the research community at large, based on the underlying citation network (diachronically).
BIP!Influence provided by BIP!
impulse
This indicator reflects the initial momentum of an article directly after its publication, based on the underlying citation network.
BIP!Impulse provided by BIP!
0
Average
Average
Average
Upload OA version
Are you the author of this publication? Upload your Open Access version to Zenodo!
It’s fast and easy, just two clicks!