An advanced artificial intelligence model, capable of contributing to the pathological analysis and diagnosis of cancer, has been created by a team of researchers, led by the Greek researcher, Marianna Rapsomaniki, from the Biomedical Data Science Center of the University of Lausanne and the University Hospital of Lausanne, in Switzerland .
Aiming to overcome the frequent hurdle of not having enough histological data to predict a patient’s cancer, the research team, co-led by Marianna Kruithof-de Julio, from the Urology Research Laboratory at the University of Bern, created the artificial intelligence model called “VirtualMultiplexer”, which produces images of biopsies and specifically the analysis of tissues through their staining, a technique used to diagnose cancer in the laboratory.
Advertising
We’ve all handled apps on our mobile phone that use a photo of us to show us what we’ll look like as an old man or as the subject of a great painter’s painting. This logic of style transfer is what this particular model utilizes, “simply instead of a painter, we have a molecular technique in the laboratory”, as Ms. Rapsomaniki characteristically explains to APE-MPE.
So using Generative AI and through its training on numerous tissue photographs analyzed in the laboratory with the application of dyes, the model creates detailed images of a cancerous tissue and the information it carries at a molecular level, an element very important for the precise identification of the disease. The results of the research were published in the journal Nature Machine Intelligence.
In the field of oncology, the technology of multiplexed imaging has appeared in recent years, where many biomarkers of cancer are detected and measured at the same time, providing greater accuracy in diagnosis, personalization of treatment and prognosis of the clinical progression of the disease. “Oncology in the field of research is experiencing a very big revolution with new techniques in the laboratory. The problem is that these techniques are very expensive and not all laboratories have the machines. So with this study we tried to achieve the same, but with a computational model,” explains Ms. Rapsomaniki.
Advertising
The goal of the researchers was with the help of this model to reduce the need to perform additional laboratory analyses, but also to be able to complement the information obtained from tissue analysis.
After building the model, the researchers tested how well these artificial images predicted clinical outcomes in order to rule out the possibility of plausible but false predictions. By comparing these predictions with real dyed tissues, the researchers confirmed that the model is reliable.
In a second stage, pathologists were given images from the lab and from the model and asked to distinguish which were artificial intelligence and which were real. It was found that the artificial images are perceived as almost identical to the real ones and this again shows the effectiveness of the model.
The researchers used tissues from people suffering from prostate cancer and pancreatic cancer. In the next stage, they aim to adapt the model so that it can be used to detect other types of cancer, but also to be used in more advanced laboratory cancer diagnosis techniques beyond tissue staining.
Although a computer engineer, Marianna Rapsomaniki has been involved for years with the use of artificial intelligence in the “fight” against cancer. “I decided quite early on because I find it shocking how complicated this disease is. The more you learn, the more you realize that we have no idea what’s going on at the molecular level or the microenvironment of the disease. It’s such a complex disease, that for me it’s very interesting to try to figure out what’s going on”, he emphasizes to APE-MPE.
As she observes, the use of artificial intelligence “has brought great hope, because these techniques can see correlations that the human eye cannot see or perceive. In addition, all the experimental techniques that are done in the laboratory give us a very large amount of data that artificial intelligence can process much more efficiently.”
“Overall,” continues Ms. Rapsomaniki, “the hope is that artificial intelligence will be a useful tool in the hands of the physician. In a context where health systems are under enormous pressure, anything that can simplify the work of doctors will help. In addition, I believe the future of oncology lies in individualized treatment for each patient, based on tumor data, because all tumors are unique, and this can only be done with advanced models.”
But, he adds, “we must not forget that these are all tools and be very careful in protecting patient data. At the same time, what is a very big research issue is to understand how these models make decisions. So we have to be sure that the results of the research can be confirmed.”
It is noted that the first author of the publication in “Nature Machine Intelligence” is the researcher Pushpak Pati, who was a member of Ms. Rapsomaniki’s team at the time of the research.
Photos and information from APE-MPE
Source: enallaktikidrasi.com
