Early detection of lung tumors; Performing effective triage in the event of a major catastrophe or improving the prevention, diagnosis and treatment of breast cancer or leukemia are just some of the applications that already reflect the present – not just the future – Use of Artificial Intelligence (AI) in Medicine.
There are many success stories that demonstrate this work, as verified this week in Madrid at the first edition of MeTechSpain, the “pinnacle” of Spanish applied research, promoted by the Federation of Technological Centers of Spain (FEDIT) It was attended by over 500 researchers, technologists and industry representatives.
A good example of this is Deeplung from Eurecat, a new system for identifying possible lung cancer suggestive nodules using AI through a tool based on deep learning. The project is carried out with the participation of the Val de Hebron Campus and is supported by the Center for Innovation in Data Technologies and Artificial Intelligence (SIDAI).
As Paula Subias, project coordinator, explains, Deeplung “originates with the aim of giving a second life to chest X-rays, which are used in hospitals for a purpose other than cancer detection, for early detection of this disease.” is done daily. It is based on machine learning algorithm which examines these X-rays and has the potential to identify early signs of lung cancer.”
It has been evaluated with a cohort of 20,000 chest X-rays and 173 patients have been identified as being at high risk of suffering from a lung nodule. 39 of them were confirmed (22.5%). After several pilot trials, it is currently undergoing validation in two hospitals. «Once tested, the use of Deeplung on unreported radiographs helps detect a monthly case that would have gone unnoticed according to current practice. This would result in around 240 cases a year in Catalonia that the health system would not be able to diagnose until at least two years later,” says Subias.
Deeplung aims to give a second life to X-rays that are made for no purpose other than tumor detection
The Bigsalud4 project has a similar objective. In this case, helping medical staff with the decision-making process, enabling better diagnosis and prognosis of diseases, and more personalized and effective treatment of patients through AI and Big Data technologies. For this, Combines data sets such as medical history, genomic information, medical imaging or life habits to build predictive models.
carried out through the Ives and Feder Fund, although it works on two specific diseases (breast cancer and acute myeloid leukemia), «the approach proposed in bigsalad4 is general enough to be able Can be applied to other tasks such as diagnosis of sepsis, prediction of readmission undefined, etc”, says François Signol, project manager and researcher in the Learning and AI line at ITI, the company that developed it.
In breast cancer, this allows automatic segmentation of dense tissue in mammograms with results similar to those performed by experts. By combining this information with epidemiological data and risk factors, a predictive model has been derived that is able to predict the presence of tumors two years from now. 73% chance that the diagnosis made to the patient is more correct, And, in leukemia, a model has been derived that predicts the risk of complications 90 days after diagnosis. «In 2019 we started this project in collaboration with different hospitals. Since then, and thanks to the progress made, it has been possible to carry out the experimental deployment of the device for breast cancer almost a year ago in collaboration with the Hospital del Mar Medical Research Institute (IMIM). And, during the following year, for a leukemia case at the Hospital La Fe in Valencia”, explains Signol.
A third example of AI healthcare applications is seen in iTriaxems, a technology to improve triage practices in the field, resource allocation, decision making, and data availability, safety and security in the context of a crisis. In events with many victims such as natural disasters or major accidents. End: reducing the number of deaths, injuries and serious consequences to those affected,
from the data of a digital sensor, placed on the wrist or other parts of the victim’s body, an algorithm processes the vital signs and establishes a treatment priority based on their severity and prognosis. This is known as triage – says Oscar González Represas, director of the ICT division of the Technological Institute of Galicia. This sensor has high-precision GPS, which allows its location at all times. In addition, victims can be located with the help of terrestrial robotics and unmanned drones equipped with artificial vision.
AI thus becomes “an additional tool”. which, in coordination with other applications and tools, will provide valuable information for decision making in said tool, but supervision is essential”, he continues. iTriaxems is in the second phase of the iProcureSecurity PCP project, an innovative European public procurement initiative Which focuses on the pre-commercial acquisition of systems for emergency medical services at the European level.
