Banca de DEFESA: José Fabrício de Carvalho Leal

Uma banca de DEFESA de MESTRADO foi cadastrada pelo programa.
STUDENT : José Fabrício de Carvalho Leal
DATE: 30/01/2024
TIME: 08:00
LOCAL: NÚCLEO DE MEDICINA TROPICAL
TITLE:

Innovation in tropical dermatology: automatic identification of cutaneous leishmaniasis lesions using artificial intelligence

KEY WORDS:

dermatology; leishmaniasis; diagnosis; AlexNet; machine learning; pictures.

PAGES: 135
BIG AREA: Ciências Biológicas
AREA: Parasitologia
SUBÁREA: Protozoologia de Parasitos
SPECIALTY: Protozoologia Parasitária Humana
SUMMARY:

Cutaneous leishmaniasis (CL) is a parasitic disease that affects approximately 1 million individuals worldwide, especially in tropical and subtropical regions. The CL polymorphism makes diagnosis difficult in health services, as the lesions can be confused with other dermatoses, such as sporotrichosis, paracocidiocomycosis and venous insufficiency. Furthermore, the diagnosis of CL depends on the experience of health professionals, such as tropical dermatologists, and laboratory procedures, making the diagnosis process slow and often late. Automated skin disease identification based on deep learning (DL) has been applied to aid diagnosis. In this study, we evaluated the performance of AlexNet, a DL algorithm, to identify images of CL lesions in patients from the Central-West region of Brazil. A set of 2458 images (up to 10 of each lesion) obtained from patients treated between 2015 and 2022 at the Dermatology Outpatient Clinic of the Hospital Universitário de Brasília were used. Of the total images, 80% were used for training, 10% for testing and 10% for validation of AlexNet in the task of identifying images of CL lesions. We performed three simulations and trained AlexNet to differentiate CL from 26 other dermatoses (e.g., chromomycosis, pyoderma, venous insufficiency). We obtained an average accuracy of 95.04% (95% Confidence Interval: 93.81–96.04), indicating excellent performance of AlexNet in identifying images of CL lesions. We conclude that automated CL identification using AlexNet has the potential to assist clinicians in diagnosing skin lesions and may contribute to the early diagnosis and timely treatment of CL. These results contribute to the development of a mobile application to assist in the diagnosis of CL in health services. Furthermore, this dissertation represents an advance in the application of DL in tropical dermatology, as a new diagnostic tool in CL.

COMMITTEE MEMBERS:
Interno - ***.733.102-** - GERSON OLIVEIRA PENNA - OUTROS
Externo ao Programa - 2676451 - CIRO MARTINS GOMES - UnBExterna à Instituição - MARCIA HUEB - UFMT