Image-to-image translation is considered a new frontier in the field of medical image analysis, with numerous potential applications. However, a large portion of recent approaches offers individualized solutions based on specialized task-specific architectures or require refinement through non-end-to-end training. In this paper, we propose a new framework, named MedGAN, for medical image-to-image translation which operates on the image level in an end-toend manner. MedGAN builds upon recent advances in the field of generative adversarial networks (GANs) by merging the adversarial framework with a new combination of non-adversarial losses. We utilize a discriminator network as a trainable feature extractor which penalizes the discrepancy between the translated medical images and the desired modalities. Moreover, style-transfer losses are utilized to match the textures and fine-structures of the desired target images to the translated images. Additionally, we present a new generator architecture, titled CasNet, which enhances the sharpness of the translated medical outputs through progressive refinement via encoder-decoder pairs. Without any application-specific modifications, we apply MedGAN on three different tasks: PET-CT translation, correction of MR motion artefacts and PET image denoising. Perceptual analysis by radiologists and quantitative evaluations illustrate that the MedGAN outperforms other existing translation approaches.

Purpose Motion is 1 extrinsic source for imaging artifacts in MRI that can strongly deteriorate image quality and, thus, impair diagnostic accuracy. In addition to involuntary physiological motion such as respiration and cardiac motion, intended and accidental patient movements can occur. Any impairment by motion artifacts can reduce the reliability and precision of the diagnosis and a motion‐free reacquisition can become time‐ and cost‐intensive. Numerous motion correction strategies have been proposed to reduce or prevent motion artifacts. These methods have in common that they need to be applied during the actual measurement procedure with a‐priori knowledge about the expected motion type and appearance. For retrospective motion correction and without the existence of any a‐priori knowledge, this problem is still challenging. Methods We propose the use of deep learning frameworks to perform retrospective motion correction in a reference‐free setting by learning from pairs of motion‐free and motion‐affected images. For this image‐to‐image translation problem, we propose and compare a variational auto encoder and generative adversarial network. Feasibility and influences of motion type and optimal architecture are investigated by blinded subjective image quality assessment and by quantitative image similarity metrics. Results We observed that generative adversarial network‐based motion correction is feasible producing near‐realistic motion‐free images as confirmed by blinded subjective image quality assessment. Generative adversarial network‐based motion correction accordingly resulted in images with high evaluation metrics (normalized root mean squared error <0.08, structural similarity index >0.8, normalized mutual information >0.9). Conclusion Deep learning‐based retrospective restoration of motion artifacts is feasible resulting in near‐realistic motion‐free images. However, the image translation task can alter or hide anatomical features and, therefore, the clinical applicability of this technique has to be evaluated in future studies.

Image-to-image translation is a new field in computer vision with multiple potential applications in the medical domain. However, for supervised image translation frameworks, co-registered datasets, paired in a pixel-wise sense, are required. This is often difficult to acquire in realistic medical scenarios. On the other hand, unsupervised translation frameworks often result in blurred translated images with unrealistic details. In this work, we propose a new unsupervised translation framework which is titled Cycle-MedGAN. The proposed framework utilizes new non-adversarial cycle losses which direct the framework to minimize the textural and perceptual discrepancies in the translated images. Qualitative and quantitative comparisons against other unsupervised translation approaches demonstrate the performance of the proposed framework for PET-CT translation and MR motion correction.