Artificial intelligence in oncology: Advancing prognosis

2024-02-02 09:34:53

Artificial intelligence tools are still making their way into the clinical practice of oncology, with patients, but they are already part of basic or laboratory research. The interpretation of millions of data and patterns will further help in understanding the response to a tumor and advance the prediction of metastasis.

Within the framework of World Cancer Day, February 4, the oncologist César Serranomember of the Board of Directors of the Spanish Society of Medical Oncology (SEOM)analyzes the expectations of artificial intelligence in the approach to cancer.

“Artificial intelligence helps manage and interpret large amounts of data, process it, and find new patterns,” says the doctor.

But it sends a clear message, “artificial intelligence in oncology is not going to replace the different specialists, it is going to be a support tool for clinical management in the different cancer processes.”

Predict the evolution of cancer

Oncology already uses prognostic tools that guide the risk of cancer reproducing and, therefore, contribute to defining the treatment before and after surgery based on the type of tumor and its aggressiveness in order to avoid relapse and metastases. .

“We have patients with a low risk of relapse and the tumor returns, but we also have patients with a high risk of relapse and the tumor never returns. Because? Because deep down biology is more complex than a series of parameters that we establish,” says the head of the Translational Sarcoma Research Group at the Vall d’Hebron Institute of Oncology (VHIO) and an expert in sarcomas at the Medical Oncology Service. from the Vall d’Hebron Hospital in Barcelona.

For that reason, “being able to identify new biomarkers or new patterns that make us better predict the behavior that cancer will have after surgery helps us and is something in which artificial intelligence will surely contribute,” he says.

“There are no tumors but patients with a tumor,” he adds, “and each day that passes, Medicine seeks to individualize and understand the response of each patient.”

The three fields of artificial intelligence in clinical practice

And the arrival of artificial intelligence in the clinical practice of oncology, according to César Serrano, is projected in three main fields: in imaging techniques; in the clinical data generated by patients; and in research.

“There is already a large implementation of artificial intelligence in preclinical research that is leading to the discovery of new molecules involved in cancer and new drugs, and we are already benefiting from it,” he explains.

Genomic sequencing, which makes it possible to identify gene mutations and identify them as biological targets against which to direct the most innovative drugs, has been one of the great steps in personalized precision medicine that seeks more effective treatments for more specific patient profiles.

“Artificial intelligence – he points out – is another step in identifying the most recurrent patterns in certain tumors. It is constantly evolving to understand and integrate thousands of data that allow us to better understand what these tumors are like and what can be most useful in treating patients.”

Oncologist César Serrano, member of the Board of Directors of the Spanish Society of Medical Oncology (SEOM), and head of the Translational Research Group on Sarcomas at the Vall d’Hebron Institute of Oncology (VHIO) in Barcelona. Photo provided

In imaging techniques

Image analysis with artificial intelligence techniques is one of the most advanced fields for use in clinical practice.

Today, the pathological anatomy of a tumor biopsy is governed by the pathologist’s assessment through not only the microscope, but also the generation of images that are digitized and that provide the name and surname of the tumors.

“Now, artificial intelligence systems are being trained to do the same as the human eye, recognize patterns” and determine the type of tumor and its characteristics, says César Serrano.

“Artificial intelligence is going to be an enormous support tool for pathologists in the future because it will facilitate the recognition of all these patterns” necessary to refine the diagnosis of cancer.

And he gives as examples, from determining the grade of a tumor to know its aggressiveness, to identifying tumor cells, as occurs in the case of cervical cytology, or even helping to rule out false positives or negatives in cases of nodules. in the lung.

In the opinion of the expert of I’m here“using artificial intelligence to diagnose cancer early today is more complicated outside of screening programs (population screening) where it can be more useful,” as is the case with the interpretation of mammograms for cancer cases. of breast.

In rare or less frequent tumors, such as sarcomas, the emergence of artificial intelligence is more complicated, a system that needs to generate new data and train algorithms and improve it with millions of patient cases.

Furthermore, another limitation is that less common cancers generate less research, so the molecular knowledge of these types of tumors is scarcer and there is less information with which to “feed” the artificial intelligence.

Artificial intelligence, fully in basic research

While in hospital consultations artificial intelligence is still taking its first steps, in basic or preclinical research laboratories, this tool is serving as support for different studies.

These are just two recent examples of research carried out by the National Center for Oncological Research (CNIO) using artificial intelligence:

• He IMPaCT_VUSCan projectco-led by CNIO and the Bellvitge Biomedical Research Institute/Catalan Institute of Oncology (IDIBELL/ICO) has been launched in order to expand knowledge about genetic variants (different versions of a gene) that affect predisposition to cancer.

Millions of genetic variants will be analyzed and with artificial intelligence Those that most influence predisposition to cancer will be sought.

The first beneficiaries of this project will be 300 families in Spain with high-risk genes that are transmitted from parents to children and in which there are therefore more cases of cancer than usual, which is considered familial cancer.

• A job of Higher Council for Scientific Research (CSIC) and the CNIO has discovered that when cancer spreads in the brain (metastasizes) it alters brain chemistry and thus interferes with neuronal communication.

The researchers measured the electrical activity of the brain of mice with and without metastasis, and observed that the electrophysiological recordings of animals with cancer are different from each other.

To ensure that this difference is attributable to metastasis, they resorted to artificial intelligence. They trained an automatic algorithm with numerous electrophysiological recordings, and indeed the model managed to identify the presence of metastases. The system even differentiated metastases from different primary tumors – skin, lung and breast cancer.

These results show that, indeed, metastasis influences brain electrical activity in a specific way, leaving a very clear and recognizable trace, something that has implications for the prevention, early diagnosis and treatment of this pathology.