After a cancer diagnosis, it is inevitable to type the type on the internet and read the first entries. Today, we go a step further: Doctor Google, now elderly, is followed by ChatGPT. But using artificial intelligence (AI) tools to confirm diagnosis and treatments is also a danger if there is no knowledge endorsed by doctors behind it.
For María García (Puerto de Santamaría) and Blanca Urbelz López-Puertas (Madrid) as 6th-year Medicine students at the Complutense University of Madrid, this catches their attention. "AI is everywhere and developing new ways to use it for the benefit of patients and doctors is necessary," says María. "We must remember that AI is not infallible and that it makes mistakes," adds Blanca.
Their mission is to train two chatbots, one to discuss esophageal cancer and the other colorectal cancer, to minimize errors and inaccuracies in providing information to two different audiences: the general population-patients and medical-researchers. "The idea is to use a different language in each case. We are clear about that," explains Blanca, who ensures that addressing patients should not necessarily mean a loss of rigor.
This idea has earned them a grant from the Cris Chair against Cancer aimed at young researchers. "This project, which will be their final degree project (TFG), actively involves students from the Complutense University in all phases as team members, from model development to answer validation, in collaboration with established researchers in the field of medical and surgical oncology," explains Julio Mayol, scientific director of the San Carlos Health Research Institute (IdISSC) and one of the research promoters.
Both students point out how Mayol's guidance has been crucial in their TFG, as well as the foundation's support. "We must never forget that without research, we do not progress," emphasizes Blanca, followed by María's request that "this activity must be encouraged. We have been fortunate that as students, we will be able to do it before others. But not everything can be just studying what exists, but continuing to advance in learning more. There is much to be done in cancer."
A Chatbot designed ad hoc to explain cancer
The students are going to design a tool that aggregates knowledge in "real-time" from all perspectives of the chosen tumors, esophagus, and colon. "Within the project, they will be responsible for tasks in prompts engineering, curation of scientific articles, and evaluation of results, promoting a multidisciplinary approach that allows them to integrate medical knowledge with artificial intelligence skills. In this way, they will gain competencies to lead future projects," details Mayol.
They emphasize that in this process, they will verify that the responses provided by the chatbot are correct. "We are going to subject it to control by specialist doctors, oncologists, who will help us verify that there are no incorrect answers," explains María. Both hope that, once completed as a TFG, it can become a useful tool ready to use. "Although it is currently just a project we are going to do as our final year project, it has the potential to become something useful that oncologists could use in their practice and recommend to patients," assures Blanca.
Mayol trusts in the potential "to evolve into a validated clinical assistant, integrable into digital health platforms." For this, the director of IdISSC describes the steps to follow: "Initially, it is developed in a closed research environment (within the Smart Health Center of IdISSC), but the possibility of commercializing it will depend on clinical validation and AI regulations in health. In any case, we want to make the best quality information accessible to patients and society equitably."
In this research, Mayol also highlights the role of Pedro Pérez Segura, Head of the Medical Oncology Service at the San Carlos Clinical Hospital, and co-author of the TFGs. At the same time, he points out the key aspects that differentiate it from other initiatives: "It is based on large language models specifically trained in oncology in our context, with advanced techniques such as Retrieval Augmented Generation (RAG) to improve the accuracy and reliability of responses in Spanish."
Regarding the technical characteristics of the project, Mayol explains that it has been evaluated and approved by the Research Ethics Committee of the institution. Of course, and although personal data will not be used, the chatbot must comply with regulations such as the General Data Protection Regulation (GDPR) and the European AI Act. Additionally, the expert emphasizes that "fundamental bioethical principles are followed: autonomy, it must provide clear and precise information, without replacing clinical judgment; beneficence and non-maleficence, risks are minimized through rigorous validations in the short, medium, and long term; and justice, it facilitates equitable access to information in Spanish, eliminating language barriers."
Artificial intelligence in Medicine, a reality or a desire?
There are precedents in the use of AI in different fields, especially in Oncology, such as decision support systems and clinical prediction models. "Furthermore, thanks to the CRIS Foundation against cancer, a Computational Oncology Unit has been launched. Although there are related international initiatives, such as OncoGPT (a collaboration between China and the USA)," Mayol explains.
Contrary to the general idea that AI is already implemented, Mayol confirms that it is indeed used in Oncology, "but its real application still faces challenges." And he lists where its use is and is not common: "It is used in image analysis, prediction of treatment responses and workflow optimization, identification of new therapeutic targets and new molecules for treatment, but it has not been fully integrated into clinical practice due to lack of validation, bias risks, and regulatory challenges."
Support for young researchers
The director of IdISSC comments on the project's impact on the students and how it helps them explore the intersection between technology, research, and healthcare management. "It allows them to acquire competencies in innovation, digital technology, artificial intelligence, and entrepreneurship in health, without forgetting the critical human component of caring for oncology patients, and prepares them to be future leaders in an excellent healthcare system based on the ethical use of artificial intelligence."
For Mayol, it is as important to instill vocation in "new researchers" as it is to "attract and retain talent, avoiding the current brain drain." In this regard, he emphasizes that these grants from the CRIS Foundation against cancer "are key because they have been created to promote the figure of the medical researcher, so that students develop a vision beyond medical care."
The CRIS Chair is under the direction of Joaquín Martínez, Head of the Hematology and Hemotherapy Service at the 12 de Octubre University Hospital and scientific director of the CRIS Foundation against cancer and Head of the Hematology and Hemotherapy Service at the Madrid center. "It is necessary for the Spanish Government to regulate the figure of the medical researcher. It is impossible to advance as a country when doctors have to conduct research in their free time."