Showing 1–1 of 1 results for author: Koschmieder, K
-
Adversarial cycle-consistent synthesis of cerebral microbleeds for data augmentation
Authors:
Khrystyna Faryna,
Kevin Koschmieder,
Marcella M. Paul,
Thomas van den Heuvel,
Anke van der Eerden,
Rashindra Manniesing,
Bram van Ginneken
Abstract:
We propose a novel framework for controllable pathological image synthesis for data augmentation. Inspired by CycleGAN, we perform cycle-consistent image-to-image translation between two domains: healthy and pathological. Guided by a semantic mask, an adversarially trained generator synthesizes pathology on a healthy image in the specified location. We demonstrate our approach on an institutional…
▽ More
We propose a novel framework for controllable pathological image synthesis for data augmentation. Inspired by CycleGAN, we perform cycle-consistent image-to-image translation between two domains: healthy and pathological. Guided by a semantic mask, an adversarially trained generator synthesizes pathology on a healthy image in the specified location. We demonstrate our approach on an institutional dataset of cerebral microbleeds in traumatic brain injury patients. We utilize synthetic images generated with our method for data augmentation in cerebral microbleeds detection. Enriching the training dataset with synthetic images exhibits the potential to increase detection performance for cerebral microbleeds in traumatic brain injury patients.
△ Less
Submitted 16 January, 2021;
originally announced January 2021.
