Gradient Pattern Analysis Applied for Computer Vision in Medical Ultrasound Diagnosis

Abstract

This paper describes a new application of the technique known as Gradient Pattern Analysis (GPA), focused here on computer vision. In the GPA domain, the image is translated into a tessellation triangulation field based on the vectors positions that make up the gradient lattice of the matrix image. The GPA version considered here generates three attributes (G1, G2 and G3) that can be used as labels for a supervised machine-learning model. The case study presented here shows that GPA is a useful tool for real-time fetal biometry from 2D ultrasound images. The application in obstetrics indicates that the technique can also be useful for learning diagnostic imaging in gynecology, hepatology and oncology. The generalization of the technique to other applications in practical learning in health is discussed.