“The exam? It needs to be redone, it’s not accurate enough!” In the future, perhaps, this phrase will no longer be heard. Thanks to Artificial Intelligence it will be possible to better manage the so-called imaging multimodale, that is the interpretation of data acquired in different ways, thus improving the personalization of the diagnosis. In addition to the degree of reliability of the tests. In short: if today it is not uncommon to undergo an echocardiography, a coronary CT scan, an MRI scan and the cardiologist asks to repeat the exam because it is inaccurate, this will no longer happen in the future, because thanks to Artificial Intelligence we are moving towards a homogenization of the quality of the examinations, their execution and interpretation, with a saving of time for the patient and economic resources for the health system. To explain it Luigi Badanofull professor of diseases of the cardiovascular system at the University of Milan Bicocca and director of the integrated cardiovascular diagnostics center at the Italian Auxological Institute of Milan, a clinical center that is part of the network of Italian IRCCS with a cardiological vocation.
The advantages of artificial intelligence techniques
One of the great expectations ofArtificial intelligence (IA) is the support to the specialist in acquiring data and images in a standardized and correct way. Echocardiography, the most prescribed examination for the diagnosis of heart disease and which uses ultrasound, is also the one most dependent on the experience of the operator who performs it, both for the acquisition of images and for their interpretation. This is precisely one of the cases in which artificial intelligence improves accuracy. “The AI - continues Badano – is able to guide the specialist during the execution of the exam, indicating for example if the image he is acquiring is not correct or inviting him to change the direction of the ultrasound probe. cases in which an underlying heart disease is detected quickly, distinguishing them from normal ones. patient”. The speed of image acquisition and their interpretation with AI is also valid for diagnostic imaging techniques in cardiology such as coronary CT and MRI, i.e. magnetic resonance. “Even the most expert of cardiologists / radiologists – explains Badano – does not exploit more than 1-2 percent of the diagnostic content of the tests performed. pixels or units of measurement of images, how images change between patients). But the potentially diagnostic content of these images is enormous. “
So the diagnosis changes with Deep learning
Manage and interpret huge amounts of data (the so-called big data), finding relationships between the information acquired to obtain a diagnosis which the cardiologist and radiologist then verify is the task of the deep learning (or deep teaching), the most innovative of the techniques. “It allows us to make previously unthinkable diagnoses and to personalize the therapy: in the case of heart failure it leads us to identify groups of homogeneous patients who, however, will respond differently to the therapies and therefore we will treat differently or who have a different prognosis. Another example of individualization of the diagnosis and treatment is the stenosis, narrowing, of the severe aortic valve, when the passage of blood is obstructed. We operated only the patients with the severe form (the surface area of the aortic valve is reduced to less than 1 centimeter square, the normality is between 2.5 and 3.5 square centimeters): today thanks to these techniques we know that even moderate aortic stenosis, with calcifications and myocardial fibrosis (presence of connective-fibrous tissue in the myocardium for which the musculature of the heart is more rigid and less contractile and the heart valves less efficient) must be regarded as severe “, spies always ga Badano.
The future that awaits us
AI techniques currently guide us in diagnosis, but the turning point in the near future will be to integrate these data with others relating to the patient’s clinical history (such as biomarkers, including genetic, proteomic and metabolomic markers that also guide us in risk assessment. cardiovascular disease) to obtain personalized information on prognosis. Proteomics (study of proteins), genomics (DNA analysis), metabolomics (study of metabolites within an organism) are all part of the “omics” sciences, which, integrated with AI, will allow medicine to take a leap and to have, for each individual patient, more and more complete information to set up truly personalized treatments. As Badano explains: “We could have for example this scenario: two patients with the same extent of valve stenosis, the first who in five years has a 70 percent chance of progressing clinically, and therefore must be kept under control; the second, with only 20 percent, on the other hand, is a patient whose controls can be reduced. All this translates into a saving of time and better service because the patient has a personalized treatment. One day AI techniques could also guide us in the diagnosis of rare heart disease “.
In collaboration with the Cardiology Network IRCCS & NExT-H