Custom built ethnic-specific dataset from SAP tests for optic pathway disease diagnosis: baseline approach using graph convolutional networks

Automated interpretation and diagnostic classification of standard automated perimetry (SAP) plots has been researched using machine learning techniques, and multiple public datasets has been proposed. Therefore, our contribution is a more complete database than the already published ones, and we pr...

Full description

Autores:: Gregory Tatis, Alvin David
Giraldo Trujillo, Luis Felipe
Marroquín, Guillermo
Jimenez Vargas, Jose Fernando

Tipo de recurso:: Trabajo de grado de pregrado

Fecha de publicación:: 2021

Institución:: Universidad de los Andes

Repositorio:: Séneca: repositorio Uniandes

Idioma:: eng

Description
Summary:	Automated interpretation and diagnostic classification of standard automated perimetry (SAP) plots has been researched using machine learning techniques, and multiple public datasets has been proposed. Therefore, our contribution is a more complete database than the already published ones, and we propose an analysis of state of the art techniques over this dataset. This ethnic-specific database, contains 66 subjects from Colombia ranging from 17-79 years old, differentiating them in two major classes: visual field (VF) of optic pathway pathological patients, and healthy eyes. Moreover, we propose, in particular, a flexible representation using Graph Convolutional Networks that combines the three SAP plots, VF indices (MD, PSD, VFI) and patient¿s data (age, pupil diameter), into a simple graph to pursue improved classification performance compared to using any single SAP plot and a variety of CNN architectures. We achieved an accuracy of 1.0 in both validation and test sets. Performance was superior compared to using other Machine Learning (ML) methods tested.

Custom built ethnic-specific dataset from SAP tests for optic pathway disease diagnosis: baseline approach using graph convolutional networks

Publicaciones similares