Machine Learning Platform for Retinography: Architecture, Engineering and Multiclass Classifier (Plataforma de Machine Learning para Retinografia: Arquitetura, Engenharia e Classificador Multiclasse)

Resumo

https://doi.org/10.5281/zenodo.20186594

This study presents the investigative path in building and evaluating machine learning models for the automatic identification of visual disorders from retinal images. The focus is strictly computational: selecting and organizing public datasets, designing reproducible preprocessing pipelines, feature extraction, choosing and comparing algorithms, and experimental validation. This is not a clinical study nor does it provide medical recommendations; the authors are not ophthalmologists. The goal is to design a platform and develop a classifier, applying principles of software engineering, data engineering/MLOps, and pattern-recognition algorithms. The experiments demonstrate the feasibility of the approach and, through comparative analysis, investigate the performance of the algorithms and their configurations, highlighting differences and similarities; the results must be interpreted within the scope of computing, not as clinical evidence or guidance.