Querying medical imaging datasets using spatial logics (Position paper)

Nowadays a plethora of health data is available for clinical and research usage. Such existing datasets can be augmented through artificial-intelligence-based methods by automatic, personalised annotations and recommendations. This huge amount of data lends itself to new usage scenarios outside the boundaries where it was created; just to give some examples: to aggregate data sources in order to make research work more relevant; to incorporate a diversity of datasets in training of Machine Learning algorithms; to support expert decisions in telemedicine. In such a context, there is a growing need for a paradigm shift towards means to interrogate medical databases in a semantically meaningful way, fulfilling privacy and legal requirements, and transparently with respect to ethical concerns. In the specific domain of Medical Imaging, in this paper we sketch a research plan devoted to the definition and implementation of query languages that can unambiguously express semantically rich queries on possibly multi-dimensional images, in a human-readable, expert-friendly and concise way. Our approach is based on querying images using Topological Spatial Logics, building upon a novel spatial model checker called VoxLogicA, to execute such queries in a fully automated way.