Assisted OCT diagnosis embedded on Raspberry Pi 4

Abstract

Machine learning in recent years has raised interest in medical imaging application, also on OCT imaging due to the straights to determine clinically significant features for diagnostics and prognostication, with potential to boost biomedical imaging interpretation and medical decision making. Evolution of hardware nowadays made possible to embedded ophthalmology imaging and machine learning techniques for a faster and precise assisted diagnosis. A downside of machine learning is the necessity of powerful hardware to compute, with latest generations of CPU or GPU to run. Low-cost effective calculation is required in this case. In this paper, successfully ported on a Raspberry Pi 4 board and reviewed machine learning algorithms to predict the presence or absence of abnormalities in the retina. A predefined dataset has been used, composed of three different diseases of the retina and a normal case of the retina. The system is portable, making it easy to be used by doctors or resident physician in their knowledge improvement.