Segmentação de Eletrocardiogramas de Cães usando Redes Neurais Convolucionais

Pedro P Santos, Nathaniel S de Oliveira, Gabriel V Paim, Luis F. N. dos Santos, Danilo Silva

DOI: 10.14209/sbrt.2025.1571149900

Evento: XLIII Simpósio Brasileiro de Telecomunicações e Processamento de Sinais (SBrT2025)

Keywords:

Abstract

Electrocardiogram exams are essential in the diagnosis of heart diseases in animals. However, manually segmenting the waves that compose an ECG is a laborious task. This work proposes an algorithm based on convolutional neural networks that accurately segments the P, QRS, and T waves in the cardiac complexes of dogs. The algorithm achieved a mean absolute error of less than two samples in 99% of predictions and a false negative rate of under 0.02%. Additionally, this proposed algorithm outperformed a wavelet transform-based method currently used in veterinary software, demonstrating performance that is approximately three times better.

Download