ComputationalMedicine 16th Oct


A key challenge in clinical medicine is the integration of omics data for the definition of a patient's status and for disease management. Computational medicine sees the “disease” in its complexity from the diagnosis to treatment and patients’ monitoring. The increasing availability of clinical, pathologic, bio-molecular and imaging data, together with the development of algorithms and techniques for data analysis and integrations, has recently led to significant advancement in the field of precision medicine.  In this seminar, we will address two key related issues: how can we effectively use the computational tools provided by Artificial Intelligence (AI) in image triage and discuss the role of multidisciplinarity in the delicate process of "making sense of data", i.e. how to move from bench to bedside and support medical decision.





What barriers remain for multidisciplinarity, L. Farina, DIAG, Sapienza University of Rome

One of the major obstacles to the introduction of precision medicine in clinical practice is the great difference in language and cultural training between those involved in data analysis ("big" or "small" whatever they are) and who is interested in defining targeted therapeutic protocols for their patients. These two worlds are not only very different from each other but, within them, they present a wide variety of approaches and conceptual frameworks. These differences, however, are so great (and increasing) that a truly multi-disciplinary work appears to be a futile effort. As a matter of fact, a dialogue is possible only when two people share some common interest. In this short seminar we will follow the example of Chauncey Gardiner, the protagonist of the film "Being there", brilliantly interpreted by Peter Sellers, and try to imagine a scenario in which the power of metaphors takes control.


How does AI help in image triage?, C. Catalano, DROP, Sapienza University of Rome

There is currently a shortage of radiologists worldwide aggravated by the increasing number of radiological exams performed. Related problems are the high number of high-cost exams that are clinically inappropriate and the consequent risk of delaying the reporting of exams that may show urgent findings. In an attempt to mitigate these issues, image triage can be performed, with the specific goals of filtering exams requests to eliminate inappropriate ones (or redirect them into a more clinically appropriate exam) and prioritizing study execution and reporting in order to favor more urgent exams.
Computational tools for image triage that are based on conventional technologies are currently available, such as clinical decision support (CDS), while prioritization of study execution and reporting is sometimes performed by radiographers and radiologists, respectively. Both approaches have several limitations and are not efficient and reliable. As a consequence, image triage is currently still not widely spread in the clinical workflow of many radiology departments.
Artificial intelligence tools and algorithms can improve the currently available approaches by automating and optimizing image triage at different levels. AI-enhanced CDS represent an advancement over conventional CDS, and could be applied to both clinical triage and prioritization of exam execution. Preliminary studies investigating the performance of deep learning algorithms in identifying urgent findings requiring prompt reporting have been published, showing exciting results.


                                            Per partecipare al webinar, è necessario collegarsi al link della stanza virtuale


© Università degli Studi di Roma "La Sapienza" - Piazzale Aldo Moro 5, 00185 Roma