New antiviral against covid with 83% effectiveness, who is it aimed at?
8 February, 2022
The Miguel Servet genetics laboratory obtains quality accreditation
8 February, 2022
Engineers from Itainnova and doctors from IIS Aragón work side by side in the design of models, based on algorithms developed with artificial intelligence, that will help identify which individuals are more likely to suffer from gastric cancer or have a worse prognosis in its evolution and need , therefore, a more detailed follow-up or early therapeutic action.
Stomach cancer is the fifth most common neoplasm and represents the third cause of death from cancer worldwide. When a doctor must assess the probability that a patient will develop gastric adenocarcinoma, the most common type, he or she knows that many factors associated with a higher risk come into play.: if there is infection Helicobacter pylori, family history, smoking, advanced age, male sex, certain genetic characteristics...
The study of the influence and degree of association of all these factors on the risk and prognosis of gastric cancer arouses great interest in the era of so-called 'precision medicine'. At the moment, Prediction models based on Artificial Intelligence (AI) and the management of big data have become very useful tools.
Currently, the 'GastricAITool' project is taking its first steps, which will put in the hands of specialists an innovative diagnostic and prognostic tool for gastric cancer based on big data technologies that will help them make decisions and detect risks. IIS Aragón (Aragón Health Research Institute) and Itainnova were chosen to carry it out within the framework of the European Gatekeeper project, funded by the Horizon 2020 program.
What do doctors expect from this tool? Regarding diagnosis, "it will allow the specialist to assess the patient's probability of developing stomach cancer and make decisions based on the results shown by the tool," explains María Asunción García González, senior researcher at the Aragonese Institute of Health Sciences ( IACS). For example, it may "recommend healthy changes in the patient's lifestyle - balanced diet, quitting smoking, regular physical activity, etc. -, prescribe treatment against Helicobacter pylori in those individuals infected by said bacteria or carry out further follow-up. detailed by gastroscopy with biopsy taking in those people in whom the tool detects a high risk of gastric cancer".
To assess the diagnostic capacity of the tool that is being developed together with Itainnova, "in the first phase the study will focus on the population group at high risk of developing this cancer: the relatives of patients with gastric cancer," says the doctor, who is a member of the Translational Research group in Digestive Pathology of the IIS Aragón and the Ciber of liver and digestive diseases.
Despite the emergence of new therapeutic options, such as immunotherapy and targeted therapies, "in general, stomach cancer has a gloomy prognosis that has not changed in recent years," he acknowledges. Less than 20% of patients survive five years after diagnosis, which in too many cases occurs at an advanced tumor stage. Other prognostic factors that influence the evolution of stomach cancer are the treatment received –mainly surgery, chemotherapy and radiotherapy–, where the tumor is located and the other diseases that the person suffers.
García González considers that "the analysis of these variables, together with the study of genetic polymorphisms involved and the subsequent development of prediction models based on artificial intelligence methodology, can provide a very helpful tool for clinical staff when making decisions." aimed at improving the prognosis of patients, either through closer follow-up or through the adoption of new therapeutic measures.
In his opinion, "prediction models based on artificial intelligence and big data have currently become very useful tools for identifying at-risk subjects and more accurately stratifying patients based on prognosis." If this tool were included in the portfolio of the National Health System, "it would be a quick, non-invasive and low-cost method that would facilitate decision-making in clinical practice, reducing costs of time, personnel and use of equipment for diagnosis and monitoring." from the patients".
The project
ANALYSIS. In this first phase, all variables will be identified and statistical analyzes will be carried out to define which are the most relevant within the framework of the project.
MODELS. In future phases of the project, models will be generated, based on algorithms developed with artificial intelligence, which will allow both diagnosing gastric cancer in early stages and predicting the prognosis of patients based on their clinical history and the study of the genes involved in this type. Of cancer.
TOOL. Once the models are generated, they will be implemented in a tool that will show doctors the patient's diagnostic and prognostic results in a simple and intuitive way. With your support in decision-making, monitoring strategies can be more personalized.
Quality data
The data that will be used to develop this project come from the results of a national multicenter study coordinated by the IACS/IIS Aragón. It includes Spanish patients and controls of both sexes and Caucasian race recruited in 16 hospitals of the National Health System between 2003 and 2008: 600 patients with primary gastric cancer (adenocarcinoma) diagnosed by gastroscopy and histological study, and 600 volunteers without gastrointestinal symptoms and free Of cancer. In addition to the follow-up data, information was collected regarding demographic data, smoking, family history of gastric cancer, comorbidities, tumor data, Helicobacter pylori infection, genomic analysis... "We have been able to gather one of the largest series of patients with gastric cancer in a Caucasian population, very well characterized clinically and with an average follow-up of 7 years –assesses Asunción García González–, so the resulting database is a magnificent tool to deepen the knowledge of such a complex and multifactorial pathology. like gastric cancer."
Explainable models
The final objective of this project is for clinical staff to have a tool to support the diagnosis and prognosis of gastric cancer. The system, based on the model generated with data from hundreds of patients, will tell the doctor if, according to their profile – clinical history, analysis, test results –, that patient is likely to develop gastric cancer or, if they already suffer from it. What is your future prognosis?
The diagnostic tool "not only will it show a result or a probability, but, thanks to explainable artificial intelligence, it will be able to show which are the most influential reasons for that diagnosis", says Vega Rodrigálvarez, project manager for the Big Data and Cognitive Systems group at Itainnova. "Evidently," he clarifies, "it is a decision-making aid tool, and it is the clinician who evaluates the diagnosis and treatment based on the results that the tool offers."
But he doesn't do it blindly. Among the challenges of this project is explainable artificial intelligence, something in high demand, "especially in the healthcare field, where experts, using the information predicted by the artificial intelligence model, can make important decisions that directly influence people. ". Rodrigálvarez recognizes that, in the health sector, "one of the reasons why AI-based tools were not very successful was the lack of transparency and explainability of the systems that were generated, the models were black boxes where a series of variables were introduced and results were offered without having any idea why that conclusion had been reached." One of the challenges of the project is to generate explainable models that, in addition to diagnosing early gastric cancer and predict a prognosis, will offer a graphical visualization that interprets and explains the results of the model, so that cancer experts can understand them. These types of techniques "will allow them to understand and trust the results that the machine has generated. ".
Source: Heraldo de Aragón
