The acquisition of large volumes of imaging data in clinical settings has prompted medical practitioners to consider disruptive technologies to efficiently process the data and extract the valuable information present in them. The medical images are used by deep learning networks to extract the clinical features for identifying the boundaries of anatomical structures or for predicting the presence of a disease. However, prior to this step, medical images need to be adequately prepared in order to be safely used and to maximise their potential in the deep learning system development. In fact supporting the development of trustworthy and resilient deep learning systems is the appropriate preparation of the medical images to be used by the deep learning solutions. Hence, many researchers have focused on the development of tools, platforms, methodologies and standards to facilitate the process of providing high-quality imaging data from clinical sites for technological advancements, while adhering with the relevant data regulations. Over the last decade, a wide range of open source image repositories and technologies have been established to promote standardised best practises for data preparation in medical imaging. There are, however, significant challenges that will require more attention and investigation. This emerging research topic aims to improve the applicability of deep learning towards real-time scenarios by increasing the trustworthiness of such methods that are currently considered as black-box.
The motivation of this special issue is to solicit the efforts and ongoing research work in the domain of data preparation for deep learning systems in medical imaging. The special issue is keen to receive articles focused on the translational medical imaging research using deep learning that are necessary to solve the data preparation problem.
Suggested topics include (but not limited to):
- Medical image de-identification (or anonymisation) for preserving patient privacy
- Synthetic data generation using adversarial networks
- Standards and guidelines for data curation in medical imaging
- Automatic detection and correction of image artefacts
- Fairness evaluation using FAIR guiding principles
- Image storage: Quality assurance workstation, archive and reading workstations
- Crowdsourcing and collaborative annotation platforms
- Cross-linkage and semantic integration of image data
- Data augmentation techniques using feasible geometric transformations, flipping, colour modification, cropping, rotation, noise injection and random erasing
- Federated learning - a new privacy preserving AI
- Ethical use of deep learning tools in medicine
- Uncertainty estimation in deep learning scenarios
