Generating and Evaluating Synthetic Medical Images
Abstract
In recent years, deep learning models have been used in various fields. Nonetheless, adaptation is still required in delicate fields like medical imaging. Since the medical field has a time constraint that necessitates the use of technology, accuracy guarantees credibility. Medical data cannot be used in machine learning applications due to privacy concerns. The development of big data is closely related to the high accuracy of disease diagnosis, classification, and recommendation. Argumentation techniques are used to expand the size of medical image datasets because deep learning-based image analysis necessitates large amounts of data. This paper proposes to include an Fréchet MedicalNet distance (FMD) value with higher accuracy than the Fréchet inception distance (FID) value in evaluating medical images that are different from the methods used to generate synthetic medical images and to calculate the FID value quickly.