Banca de DEFESA: ADRIELLY SAYONARA DE OLIVEIRA SILVA

Uma banca de DEFESA de MESTRADO foi cadastrada pelo programa.
STUDENT : ADRIELLY SAYONARA DE OLIVEIRA SILVA
DATE: 22/10/2024
TIME: 15:00
LOCAL: Laboratório Didático de Computação Biomédica
TITLE:

Machine Learning Application on EEG Signals for the Diagnosis of Autism Spectrum Disorder

KEY WORDS:

Autism Spectrum Disorder (ASD); early diagnosis; Random Forest; Electroencephalogram (EEG); machine learning.

PAGES: 60
BIG AREA: Engenharias
AREA: Engenharia Biomédica
SUMMARY:

Autism Spectrum Disorder (ASD) is a complex and heterogeneous neurological condition, with an estimated prevalence of approximately 1 in 44 children. Early diagnosis is essential to optimize the quality of life of affected individuals, as it allows the implementation of effective therapeutic interventions during critical periods of child development. This study aims to develop an early differential diagnosis based on electroencephalogram (EEG) signals, with the aim of identifying features associated with ASD. For this purpose, EEG signals from 56 individuals extracted from the Sheffield database were used, applying techniques for the treatment of missing data. Subsequently, a manual selection of the EEG data was performed to suit the analyses. Several machine learning methods were employed, resulting in high classification performance for the two analyzed datasets: one with 9 electrodes and the other with 15 electrodes. Preliminary results indicate that the analysis of EEG signals with configurations of 9 and 15 electrodes has significant potential for the identification of patterns associated with ASD. In particular, the Random Forest model with 500 trees stood out in both datasets, achieving an accuracy of 98.06% in the 9-electrode dataset and 98.49% in the 15-electrode dataset. These findings suggest that the proposed model may serve as a promising tool to support the clinical diagnosis of ASD, providing a faster and more accurate analysis of brain signals. The robustness and effectiveness observed in the models highlight the feasibility of using EEG combined with machine learning techniques to improve the early diagnosis of ASD.

COMMITTEE MEMBERS:
Interna - 2727505 - GISELLE MACHADO MAGALHAES MORENO
Presidente - ***.618.664-** - JULIANA CARNEIRO GOMES - UFPE
Externa à Instituição - MAIRA ARAUJO DE SANTANA - UFPE
Interno - 1807632 - WELLINGTON PINHEIRO DOS SANTOS