Cancer treatment is a long-time process that is not only confined in hospitals and clinics. In the past few years, a trend to patient-centric frameworks and approaches in cancer treatment has prevailed. Answering these needs, current technology developments, such as Artificial Intelligence, are challenging the Iron Triangle of Health concept and holding the promise that health care access and quality can be improved while reduction costs.
In this process, ASCAPE aims at disrupting the model paradigm by developing AI models that will improve the Quality of Life (QoL) of cancer survivors while reducing costs to the healthcare systems and improving access to its services. This will be done by collecting input from patients and their devices directly, leading to increased AI results and quality of care, and decreasing the administrative effort for collecting data, lowering the cost of care and improving access by freeing time and human resources. During the project, ASCAPE will focus on the training of this AI in two types of cancer: breast and prostate.
ASCAPE’s goal
The main goal for ASCAPE is to create an open AI infrastructure that will enable health providers, such as hospitals, research institutions and companies, to deploy an edge node locally and execute the AI algorithms on their private data without the need to send their data outside their domain. The data-derived knowledge will be made available to doctors to aid them in their decisions and help provide a better Quality of Life for their patients. ASCAPE’s long-term vision is to create a framework where multiple healthcare stakeholders will work together and share knowledge about various types of cancer and, if possible, other kinds of health issues.
ASCAPE’s three main pillars
ASCAPE builds its value proposition around three pillars:
- Data-driven Machine Learning-based as enabler for Personalized Healthcare: An ASCAPE Edge Node will take advantage of AI analytics on patients’ data and predictive models to obtain QoL risk predictions and suggestions of interventions.
- Continuous Data Collection and Learning: Data stemming from the process mentioned above will re-feed the ML model to further improve the quality of interventions.
- Cooperative, Distributed, Open: ASCAPE solution is planned to foster healthcare stakeholders to share knowledge.
ASCAPE’s novel model training federated learning can support the dynamic reconfiguration of the federation and underpin continuous learning from patient data updates. Moreover, machine learning algorithms on homomorphically encrypted datasets is being introduced in order to strengthen patients’ privacy.
Furthermore, to prevent a disclosure of personal data four factors were introduced: a two-level de-identification, differential privacy to prevent disclosure from Machine Learning models, a federated distributed learning where unencrypted patient data remain at HIS sites only, and model training over homomorphically encrypted patient data on the ASCAPE cloud.