Evolution of conditional-GANs for the synthesis of chest X-ray images
Por:
Rodriguez-de-la-Cruz, Juan-Antonio, Acosta-Mesa, Hector-Gabriel, Mezura-Montes, Efren, Ardmbula Cosio, Fernando, Escalante-Ramirez, Boris, Olveres Montiel, Jimena
Publicada:
1 ene 2021
Resumen:
Deep learning (DL) is now widely used to perform tasks involving the
analysis of biomedical imaging. However, the small amounts available of
annotated examples of these types of images make it difficult to use
DL-based systems, since large amounts of data are required for adequate
generalization and performance. For this reason, in recent years,
Generative Adversarial Networks (GANs) have been used to obtain
synthetic images that artificially increase the amount available.
Despite this, the usual training instability in GANs, in addition to
their empirical design, does not always allow for high-quality results.
Through the neuroevolution of GANs it has been possible to reduce these
problems, but many of these works use benchmark datasets with thousands
of images, a scenario that does not reflect the real conditions of cases
in which it is necessary to increase the data due to the limited amount
available. In this work is presented cDCGAN-PSO, an algorithm for the
neuroevolution of conditional-GANs (cGAN) that adapts the concepts of a
previously reported neuroevolutionary algorithm called DCGAN-PSO, which
was focused on the design and training of DCGANs through the use of
Particle Swarm Optimization, a Swarm Intelligence algorithm that uses a
set of potential solutions to approximate a highly competitive solution.
The evolved cGANs allows the synthesis of three classes of chest X-ray
images and they were trained with only 600 images of each class. The
synthetic images obtained of each class show good similarity with real
chest X-ray images.
Filiaciones:
Rodriguez-de-la-Cruz, Juan-Antonio:
Univ Veracruz, Inst Invest Inteligencia Artificial, Campus Sur,Paseo 112, Xalapa 91097, Veracruz, Mexico
Acosta-Mesa, Hector-Gabriel:
Univ Veracruz, Inst Invest Inteligencia Artificial, Campus Sur,Paseo 112, Xalapa 91097, Veracruz, Mexico
Mezura-Montes, Efren:
Univ Veracruz, Inst Invest Inteligencia Artificial, Campus Sur,Paseo 112, Xalapa 91097, Veracruz, Mexico
Ardmbula Cosio, Fernando:
Univ Nacl Autonoma Mexico, Inst Invest Matemdt Aplicadas & Sistemas, Campus Sur,Paseo 112, Xalapa 91097, Veracruz, Mexico
Escalante-Ramirez, Boris:
Univ Nacl Autonoma Mexico, Fac Ingn, Univ 3000,Ciudad Univ, Cd Mx 04510, Mexico
Olveres Montiel, Jimena:
Univ Nacl Autonoma Mexico, Fac Ingn, Univ 3000,Ciudad Univ, Cd Mx 04510, Mexico
|