Gastric cancer is a high-impact disease that affects millions of people worldwide. With its high incidence and poor prognosis, it has become the fifth most common cancer and the third leading cause of cancer-related deaths. The complexity of this disease lies in the interaction of multiple factors, such as demographic, clinical, environmental factors, Helicobacter pylori (H. pylori) infection, and genetic factors.
In light of this reality, it is crucial to have advanced tools that allow for early diagnosis and prediction of adverse outcomes. In this regard, the GastricAITool project has emerged as an intelligent solution based on Artificial Intelligence and Big Data technologies. Its goal is to provide clinicians with the necessary support in making crucial decisions and facilitate personalized treatment strategies for patients. The GastricAITool project has been developed by the Technological Institute of Aragon (ITAINNOVA) in collaboration with the Aragon Institute of Health Research (IISA).
How was GastricAITool built?
In order to develop the GastricAITool project, a database obtained after the execution of a multicentre, national case-control study coordinated by IISA in which 16 hospitals from 7 regions in Spain participated. 600 patients with primary gastric adenocarcinoma and 600 healthy individuals without cancer were included in the study. This database contains a large volume of demographic data (sex, age, smoking habits), clinical data (family history of gastric cancer, pathological history, comorbidities, treatment received, Helicobacter pylori infection status, overall and disease-free survival), tumour data (location, histological type, staging, development of metastasis) and genetic data (genotyping of a panel of 250 polymorphisms). Thus, the database generated represents one of the largest series of patients with gastric cancer described in Europe, very well characterised clinically and with an average follow-up of 7 years, making it a magnificent «tool» for deepening our knowledge of a pathology as complex and multifactorial as gastric cancer.
In the development of GastricAITool, statistical and artificial intelligence techniques were employed to select the most relevant variables and build diagnostic and prognostic models. Special attention has been given to exploring Explainable Artificial Intelligence methods, such as Shapley values, to provide interpretability of the models. Explainable Artificial Intelligence offers the possibility of using complex or opaque models, commonly known as «black boxes», without sacrificing transparency in the process. This approach is crucial for clinicians as it allows them to understand how the results were obtained and, consequently, make informed clinical decisions.
For diagnosis, significant risk factors or variables were selected, such as H. pylori infection, tobacco consumption, family history of gastric cancer, and 16 genetic variables located in genes primarily related to immune response control and DNA repair processes. As for prognosis, the variables used in the model included cancer stage, presence of metastasis, tumour location, treatment received (chemotherapy, radiotherapy, or surgery), gender, and patient age.
The final diagnostic and prognostic models, which constitute the main engine, were integrated into the final tool.
How to use GastricAITool?
The functionality of GastricAITool is presented in an intuitive and user-friendly interface. The main steps to use it are as follows:
- Clinicians or researchers can access the tool through the main page, where they are prompted to enter their username and password.
GastricAITool Main Page
- Once logged in, the patient list viewer is displayed, where they can edit, view existing data, or add new patients to predict diagnosis or prognosis using the integrated models.
- When entering a new patient, the patient’s information necessary for model application is requested.
- Once the information is completed and saved, the user can click on the «Save & Execute» button. The model result is then displayed along with explanatory graphs illustrating the importance of the variables and providing an interpretation of how the model arrived at those results.
Example of prognosis results for a patient
Example of global interpretability of the prognosis model
To conclude, we would like to share some final reflections:
- It is essential to use Big Data and Artificial Intelligence techniques and tools for a positive purpose, such as improving the health of citizens. These technologies have great potential to drive significant advances in the medical field.
- The application of artificial intelligence must go hand in hand with experience and contextual knowledge. In this regard, the collaboration between the Big Data and Artificial Intelligence group at the Technological Institute of Aragon (ITAINNOVA) and the Translational Research Group in Digestive Pathology at the Aragon Institute of Health Research (IISA) has been crucial in achieving the results obtained. The combination of specialized knowledge in both areas has been key to the success of the project.
- Participating in the European project GATEKEEPER, which aims to improve the quality of life of citizens through the provision of health services, has been a personally rewarding experience. Our dream is to contribute to a healthier and happier world, and being part of this project brings us closer to that goal.
In summary, the responsible use of Big Data and Artificial Intelligence technologies, in collaboration with experts in the medical field, can have a significant impact on improving health and quality of life. We are excited to continue working on this path and contribute to a more promising future in the field of health.
Get in touch!
If you are interested in the Gatekeeper project, stay tuned through our social networks, webinars and newsletter.
Follow the project:
- Gatekeeper Project Website
- Gatekeeper LinkedIn Group
- Gatekeeper Twitter Account (@Gatekeeper_EU)
- Gatekeeper YouTube Channel
To get in touch with ITAINNOVA, IISA and the project team: