The days where all our data was hosted on internal local servers are gone, with cloud computing becoming the way of the future, revolutionising the way that organisations deal with the storage of high volumes of data. However, despite the numerous benefits that this technology can provide (e.g. flexibility, low investment, less operational issues, availability, and easy implementation), less than 10% of the world's data is currently stored in the cloud. Security and privacy are still the biggest concerns that organisations indicate with respect to cloud adoption. The idea of storing data externally, often at centres located in remote areas of the world, is a scenario that doesn’t help address these misgivings.
ATMOSPHERE is working on increasing confidence on cloud to encourage different business sectors to take up this technology and ultimately improve business efficiency and competitiveness.
The demand for healthcare will continue to rise, mostly due to population aging and growth, but also due to an increasing concern for wellbeing among patients. It's expected that these forces will influence the industry role in IT and, by association, that of cloud computing. Health domain data is naturally bound by privacy restrictions, despite that it can be released for research under specific conditions (e.g. general interest, health management, research with informed consent). Total anonymisation is sometimes difficult to apply without affecting data (e.g. in cranial CT) or even impossible (e.g. in genomic data). Access should be restricted, and derived data may be even more restrictive (e.g. genomic variant calling, cranial segmentation). Finally, data should be processed timely and availability is a must. With trustworthy cloud computing levels, healthcare organisations and providers can share data securely, both internally and externally, manage privileged users, and comply with monitoring and reporting regulations.
In addition, it will contribute to the paradigm shift from conventional, qualitative radiology reading to quantitative patient evaluation, helping expand precision medicine.
The transference and remote processing of medical imaging and its associated health data are perfect examples of services that require a high degree of trustworthiness. Trustworthiness dimensions such as performance, availability, privacy preservation, ethical management, security, stability, and provenance are very important to ensure fulfilling legal and ethical frameworks (especially important for the EU-GDPR) and healthcare provisioning.
A Medical Imaging Biomarker (MIB) is a computational process over digital images that obtains a quantitative feature of a physiological, chemical, histological, anatomical, physical or metabolic behaviour. An Image Biomarker consolidates statistical values and produces parametric images from features obtained from the whole image or from a region of interest selected by the physician. Image Biomarkers are based on physical and data analytics models that are validated on large cohorts and are applied to specific patient data. On one hand, tuning and training image biomarkers require high computational effort, and on the other hand, the projection or application of the model over specific patient data may require less resources.
The availability of portable and affordable medical imaging devices has facilitated its widespread and massive use in routine and population screening. However, interpretation is complex and there is a lack of enough skilled professionals that could either travel or go through massive image acquisition programs. The use of biomarkers and automatic analysis techniques for medical images can facilitate the stratification of the cases, so patients susceptible of specific diseases can be processed first or receive more attention. Imaging biomarkers are quantitative values relevant to a patient's diagnosis or treatment follow-up. By defining specific targets, imaging biomarkers can provide parametric images or indicators that could support clinical decision.
The project will demonstrate the benefits of the technologies to assess and improve the trustworthiness of health management services in the context of the analysis of echocardiographic images. The analysis of textures, for example, could be a good surrogate indicator for fibrosis, which is also correlated to other diseases and of course to the risk of cardiac failure. The processing will require intensive computing resources, high availability, adaptation to workload, privacy preservation of data and transmissions, robustness of the services, and other features, which need to be assessed both a priori and dynamically during the execution of the service.