Meeting Banner
Abstract #2820

Augmentation of Diffusion-Encoding Gradient Directions with Deep Learning is not suitable for dMRI clinical studies

Justino Rafael Rodríguez-Galván1, Carmen Martín-Martín1, Antonio Tristán-Vega1, Carlos Alberola-López1, and Santiago Aja-Fernández1
1Laboratorio de Procesado de Imagen, Universidad de Valladolid, Valladolid, Spain


Validation in Deep Learning for enhancement of diffusion Magnetic Resonance Imaging results usually sticks to conventional image similarity metrics. Despite those results, further research on synthetic data may result in discordance with the real one. In this paper we have compared 61 real gradient directions against 61 quasi-identical synthetic gradient directions, obtained by subsampling the real ones, for the assessment of the differences between chronic and episodic migraine patients. Even with high image comparison metrics, differences in t-test are not compelling. For that reason, we do not recommend synthetic images for clinical use.

This abstract and the presentation materials are available to members only; a login is required.

Join Here