One of the most promising areas of application for artificial intelligence is healthcare. This technology's ability to analyze large amounts of data and learn from it allows it to accelerate the diagnosis of certain tumors, with promising future uses, for example in mammograms. AI not only has the potential to make an important contribution to blood donations, by improving communication with potential donors and identifying in advance which hospitals will need more blood or plasma based on the location of the donors. donations and serious incidents. In this article, in collaboration with professors from the IMT Lucca School of Advanced Studies Ennio Bilancini and Massimo Riccaboni, we analyze in detail these two examples of application of artificial intelligence to healthcare to fully understand the potential and the innovations that this technology already brings. Italian and world hospitals.
The challenge: increase blood donations using AI
According to data updated until 2022 from Avis, donations in Italy reached 2.8 million, a slight increase compared to 2021. The number of donors, announced by the National Blood Center, reached 1,660. 227 the same year. Slight growth which allows Italy to be self-sufficient in the management of transfusions. What is most concerning is the lack of generational renewal among donors. In fact, the percentage of donors aged 18 to 45 continues to decline. The collection of plasma – the liquid part of blood – remains critical and ended in 2022 with insufficient quotas to produce the plasma-derived medicines needed for Italian patients.
This is an image that could be improved thanks to the intervention of artificial intelligence. As Ennio Bilancini, professor of economics at the IMT School of Advanced Studies in Lucca and specialized in behavioral economics, explains: “It is important to identify who will be most inclined to donate and when they will be most likely to donate. inclined to donate. Being able to communicate with donors is also very important, as donors do not have all the information about when it is important to donate and how often. And AI can help achieve this goal in the near future.
Artificial intelligence to anticipate demand
“Artificial intelligence can allow us to identify and predict the extent of demand in certain places and at certain times. For example, understanding how much type 0 blood will be needed at the hospital in a certain city on a given day,” Bilancini continues. Once programmed, artificial intelligence systems learn autonomously to use and interpret the data provided to them. By cross-referencing the causes leading to the need for transfusions with the availability of bags in blood banks and many other data, AI would theoretically be able to predict with precision – until a few years ago this was unthinkable – the quantity of blood that will be injected. necessary and where. This would allow us to optimize the distribution of resources, ensuring that blood bags are available when and where they are most needed.
The contribution of AI to improve communication
But predicting demand and guaranteeing the supply of blood already collected is not enough. It is also essential to communicate effectively with blood donors and recipients. And here too, artificial intelligence comes into play. Through advanced analytics, it is possible to better understand donor behaviors and develop personalized communication strategies for each demographic, interest group and age. “It’s important to remind donors when it’s time to donate and how often,” Bilancini emphasizes, “and artificial intelligence can help us determine the best way to do so.” In fact, AI makes it possible to “identify who will be most inclined to donate and when they will be most inclined to do so,” concludes Bilancini.
The AI revolution in breast cancer diagnosis
Apart from blood transfusions, another crucial area where AI can revolutionize the healthcare sector is in the early detection of breast cancer. Every year, 56,000 Italian women are diagnosed with this type of tumor, but in 2022, 2,000 of them were diagnosed late, also due to the delays accumulated during the pandemic. A situation that can be improved thanks to artificial intelligence, which in these cases could literally help save lives.
Currently, breast cancer screening – which is carried out using a special x-ray called a mammogram – is a costly procedure subject to limitations, regularly offered, according to ministerial indications, to women who are already over fifty in order to reduce their exposure to radiation. Today's devices emit very little radiation, but this continues to be a fear that causes many women not to seek examinations, even though the minimal harm caused by radiation is negligible compared to the benefits.
Artificial intelligence for early and accurate diagnosis of breast cancer
It is therefore essential to increase the accuracy and speed of diagnoses in order to be able to carry them out more, more precisely, and also to convince women who still do not have confidence in themselves to carry out regular check-ups. Massimo Riccaboni, professor of economics at the IMT School of Advanced Studies in Lucca specializing in economic studies in the health sector, explains that “we need to have new technologies that allow wider access to screening” . Speeding up mammograms and making them more accessible across the country would reduce waiting lists, giving more women the opportunity to control the development of any tumor masses. The Mamoscreen project, currently under development, is an example of how these results can be achieved using artificial intelligence.
Mammograms with artificial intelligence
Having thousands and thousands of mammograms to train on, the project intends to use AI to know in which cases the images show a benign tumor, in which cases it is malignant and in which cases there is no to worry. The more you use it, the more the precision increases, which is however not absolute. This is why we cannot ignore the control of human specialists, who are, however, aided by the new systems. “Using artificial intelligence allows us to increase diagnostic accuracy and reduce costs because it reduces the time spent by healthcare workers, which is a scarce resource. Additionally, diagnostic accuracy also increases because the system learns globally from subjects who pass the screening,” says Riccaboni.
Artificial intelligence for tumor diagnosis
This is not the only case in which artificial intelligence can be used for cancer diagnosis. Among the most difficult oncology cases to treat is pancreatic cancer, which, thanks to AI analysis of cases already present in families and patients' lifestyle habits, can be diagnosed up to three years earlier than a few months ago. Similar devices were also installed at the Gemelli Polyclinic in Rome in early 2023 for the diagnosis of gynecological tumors. Specifically, the algorithms are able to predict rapidly changing cellular development. Although attention to artificial intelligence has increased exponentially in conjunction with the release of ChatGPT in late 2022, assisted diagnosis systems have been in development well before that. Already in 2020, for example, it was predicted that artificial intelligence would soon reach the accuracy of human diagnostics.
The next episode
What you have just read is the third episode of Pillole di Scienza, the Open series in collaboration with the IMT Lucca School of Advanced Studies. In the first episode, we analyzed how blockchain works, the opportunities offered by this technology and how to recognize scams that exploit it to defend themselves. In the second, we debunked seven myths about hackers and cybersecurity. We are now halfway through the series. Stay tuned for the next three episodes. One per week, always in Open.
Ennio Bilancini is Professor of Economics and PhD Coordinator in Economics, Analysis and Decision Sciences at IMT School of Advanced Studies Lucca
Massimo Riccaboni is professor of economics and director of the research unit Axes (Laboratory for Analysis of Complex Economic Systems) of the IMT School of Advanced Studies Lucca