ZELUS and Stockholm University begun collaborating on an exciting new joint research venture aiming to enhance the care provided to skin cancer patients. This project, ASME (Artificial intelligence Supporting CAncer Patients across Europe), is a sub-project of the Horizon 2020 project ASCAPE (Artificial intelligence Supporting CAncer Patients across Europe), funded by the European Commission.
ASME will focus on markers that can predict the possibility of developing side effects as well as progression of the melanoma during adjuvant therapy. Work will be based on a Karolinska University Hospital (KIH) retrospective dataset including data from a cohort of melanoma patients that received adjuvant therapy in the period between 2019 and 2021.
Two streams of development are foreseen: one will involve the creation of new algorithms to analyse and make use of the dataset (and future extended versions both from KIH and other hospitals) for ML-based personalized outcome predictions and intervention suggestions, whereas the other will involve extending the ASCAPE prototype (which also provides personalized outcome predictions and intervention suggestions) so that it supports melanoma patients on the basis of the KIH dataset’s data model translated into ASCAPE/FHIR7 standards. These two lines of development will meet at the point of evaluation, where the two AI solutions will be examined independently and in comparison, to each other, both analytically and from a clinician’s perspective. Moreover, ASME will provide an external evaluation of ASCAPE’s open AI Infrastructure with respect to (1) the ease with which existing algorithms can be extended to a new type of cancer and (2) the ease with which the open AI Infrastructure can be extended with new algorithms.
With regards to the sub-call’s focus areas, ASME will extend existing ASCAPE datasets with an ASCAPE-ready melanoma patient dataset to enable (both standard ASCAPE and ASME-proposed) AI/ML algorithms (for model training, predictions, analytics and simulations) on a type of cancer not originally supported by ASCAPE, without sighing away from proposing alternative algorithms training/prediction and explainability algorithms and visualizations and comparing them to ASCAPE’s to find the optimum candidates for integration with ASCAPE.
ASME’s approach is to both make use of, showcase ASCAPE’s strengths and openness, build on it to provide a clinical trial ready ASME/ASCAPE integrated prototype and enhance it where appropriate.