A carregar...
Publicação
Applications of artificial intelligence to pattern recognition in the field of skin cancer : state of the art and what to expect from the future
| Nome: | Descrição: | Tamanho: | Formato: | |
|---|---|---|---|---|
| 2.34 MB | Adobe PDF |
Orientador(es)
Resumo(s)
O sistema de saúde está a testemunhar o surgimento da inteligência artificial na área da dermatologia, aprimorando a tomada de decisão clínica personalizada. Ao longo dos anos, a incidência de melanoma aumentou nos países desenvolvidos, principalmente nos de etnias de pele clara. De forma geral, a incidência está a aumentar devido à exposição a radiação ultravioleta e ao aumento da esperança média de vida. Se detetado numa fase inicial, o melanoma é altamente tratável, tornando a deteção precoce crucial. Portanto, a tecnologia de inteligência artificial revela um grande potencial para melhorar o diagnóstico precoce do melanoma, reduzindo o número de sobrediagnóstico e subdiagnóstico de lesões cutâneas pigmentadas suspeitas. O objetivo desta revisão é explorar a combinação de inteligência artificial e humana no diagnóstico de melanoma. Desta forma, pretendo apresentar uma introdução básica dos conceitos de inteligência artificial, bem como uma visão geral da pesquisa atual de aprendizagem profunda no diagnóstico de melanoma, comparando a performance dos clínicos e os algoritmos de aprendizagem profunda. Avaliarei criticamente os estudos que avaliam a efetividade dos métodos e discutirei limitações e desafios futuros. Literatura relevante e artigos científicos foram selecionados para esta revisão. Inclui estudos que utilizam aprendizagem profunda, especificamente redes neurais convolucionais profundas. Estes últimos mostraram a sua eficácia para a classificação de bases de dados de imagens. Com base na revisão da literatura, defendo que os sistemas de diagnóstico por computador no contexto do diagnóstico de melanoma representam um desenvolvimento futuro significativo para a prática clínica diária da dermatologia.
The healthcare system is experiencing an upsurge of artificial intelligence in dermatology, improving personalized clinical decision making. Over the years, the incidence of melanoma has increased in developed, mostly fair-skinned ethnicities. This increase is mainly due to exposure to ultraviolet radiation and increasing life expectancy. If detected at an early stage, melanoma is highly treatable, so early detection is critical. Therefore, artificial intelligence technology offers great potential to improve the accuracy of early melanoma detection and reduce the number of overdiagnosis and underdiagnosis of suspicious pigmented skin lesions. The aim of this review is to examine the combination of human and artificial intelligence in the diagnosis of melanoma. To this end, I will provide a basic introduction to the concepts of artificial intelligence and review current research of deep learning in melanoma diagnosis by comparing the performance of physicians and deep learning algorithms. I will critically evaluate the studies that assess the effectiveness of these methods and discuss limitations and future challenges. Relevant literature and scientific articles were selected for this analysis. These include studies that use deep learning, particularly deep convolutional neural networks. The latter have demonstrated their effectiveness in classifying image data. Based on the literature review, I argue that computer-aided diagnosis systems related to melanoma diagnosis represent a significant future development for the daily practice of dermatology.
The healthcare system is experiencing an upsurge of artificial intelligence in dermatology, improving personalized clinical decision making. Over the years, the incidence of melanoma has increased in developed, mostly fair-skinned ethnicities. This increase is mainly due to exposure to ultraviolet radiation and increasing life expectancy. If detected at an early stage, melanoma is highly treatable, so early detection is critical. Therefore, artificial intelligence technology offers great potential to improve the accuracy of early melanoma detection and reduce the number of overdiagnosis and underdiagnosis of suspicious pigmented skin lesions. The aim of this review is to examine the combination of human and artificial intelligence in the diagnosis of melanoma. To this end, I will provide a basic introduction to the concepts of artificial intelligence and review current research of deep learning in melanoma diagnosis by comparing the performance of physicians and deep learning algorithms. I will critically evaluate the studies that assess the effectiveness of these methods and discuss limitations and future challenges. Relevant literature and scientific articles were selected for this analysis. These include studies that use deep learning, particularly deep convolutional neural networks. The latter have demonstrated their effectiveness in classifying image data. Based on the literature review, I argue that computer-aided diagnosis systems related to melanoma diagnosis represent a significant future development for the daily practice of dermatology.
Descrição
Trabalho Final do Curso de Mestrado Integrado em Medicina, Faculdade de Medicina, Universidade de Lisboa, 2022
Palavras-chave
Inteligência artificial Aprendizagem profunda Cancro da pele Melanoma