Teresa Bau
Artificial intelligence is one of the most promising technologies in the field of health research. Two projects that use artificial intelligence technologies in two very different fields have been selected in this year's eHealth Center From idea to project awards: improving patient care, and research into drugs for metastatic cancer. The prizes are endowed with €3,000 for master's degree final project in digital health by students or graduates of the Faculty of Health Sciences and the Faculty of Computer Science, Multimedia and Telecommunications.
Sarai Suárez, an expert with a bachelor's degree in Audiovisual Communication who is studying the Master's Degree in E-Health at the Universitat Oberta de Catalunya (UOC), has won the award with the project "Improving doctor-patient communication with AI", which aims to make the medical reports received by patients easier to understand in order to promote a more effective dialogue between the two parties and active participation by patients in making decisions about their health. "Not understanding medical terms, the diagnosis, the prescribed treatment or the recommendations made when discharged can cause poor adherence to treatment, an inability to recognize warning signs and poor follow-up of visits", explained Suárez, who has carried out her project within the framework of her experience as a nursing assistant at Vall d'Hebron Hospital, where she was able to learn more about the world of healthcare.
Suárez's project, which she is carrying out at Vall d'Hebron Hospital, involves testing three models of generative artificial intelligence technologies that produce simplified and enriched versions of medical reports in accordance with the patient's level of health literacy, with easy-to-understand explanations of the medical terms and abbreviations used. The aim is to facilitate the reading and understanding of these documents and see if this technology can be used in clinical practice in the future.
The project assesses three different generative AI technologies: Meta's open-source model Llama 2 and two proprietary code models, ChatGPT and Claude by Anthropic. The results will help create a roadmap to develop a study protocol with patients.
"AI allows you to take complex texts and quickly and automatically 'translate' them into simplified versions with a vocabulary adapted to the patient's knowledge. Unlike previous rule-based solutions, AI learns these simplifications directly from real-life human language examples. Another advantage is that this translation work does not require extra time from the medical team," she said.
If the preliminary results of this pilot are positive, which for now they are, a study with patients in a testing environment that does not affect the care activity at the hospital will be considered. If successful, this technology could be implemented in standard clinical practice in the future.
��Sarai Suárez, alumni del máster #SalutDigital @uocesalut: “El uso de IA puede mejorar la comprensión de los informes médicos y favorecer la participación activa del paciente y la alfabetización en salud”, ganadora de 'De la Idea al Proyecto' @eHealthUOChttps://t.co/FnGIvO07qb pic.twitter.com/o2DDPmtmjo
— eHealth Center UOC (@eHealthUOC) February 22, 2024
In-silico discovery of new drugs for metastatic cancer
The project by Maria Butjosa, who holds a bachelor's degree in Biochemistry from the Universitat Autònoma de Barcelona (UAB) and a Master's Degree in Bioinformatics and Biostatistics from the Faculty of Computer Science, Multimedia and Telecommunications of the UOC, seeks to discover new drugs for the treatment of metastatic cancer, a field in which there is a significant shortage and where patient treatment protocols are lacking. Her internship in the Cancer Computational Biology Group at the Vall d'Hebron Institute of Oncology (VHIO), carried out under the supervision of Dr José A. Seoane, inspired her to focus on this 'pending issue' in the treatment of cancer.
The "In-silico discovery of drug response differences between primary and metastatic cell lines" project involves creating and validating a methodology for the identification of drugs with an impact on the elimination of cells not just of metastatic cancers but also of primary cancers. To achieve this, she uses public databases containing the results of studies of thousands of combinations between cells of different types of cancers and potential drugs. Deep learning algorithms will also be used to predict the response of drugs to certain types of cells for combinations that have not been studied. Butjosa explained that the innovation "comes from studying the response of drugs in differentiated cells, depending on whether they're from primary or metastatic cancers, instead of from the tumour's gene signature as is usually done. The project also studies the action of drugs grouped according to their mechanism of action in addition to studying them individually."
Machine learning technology provides many possibilities for cancer research: it allows us to discover biological relationships and make predictions about the outcome of treatments that it would otherwise not be possible to obtain. She highlighted that "current facilities for analysing tumour DNA, with the generation of large volumes of data that this entails, allow machine learning to be used to analyse information and reach valuable conclusions". The challenge is how to incorporate this technology, which is currently being used at a preclinical level, into daily clinical practice, where it can have great potential.
Butjosa's project also has the added value of incorporating the gender perspective into research: "in the past, gender and sex bias in the world of medical research has meant that innovations weren't applicable to all patients, although this has changed in recent years and is now being taken into account when designing studies," she said.
The research is currently at the stage of applying the methodology developed by Butjosa to various databases, and the next step will be to optimize the deep learning algorithm. If the results are positive and she finds any potentially effective drugs or groups of drugs, they will be tested in cell cultures or on mice in the laboratory.
��Maria Butjosa, alumni del máster #Bioinformática y #bioestadística @uoceimt: “El deep learning puede optimizar el tratamiento de los pacientes con cáncer metastásico”, ganadora de 'De la Idea al Proyecto' @eHealthUOChttps://t.co/cXkSVWxeZH pic.twitter.com/kON1arCemi
— eHealth Center UOC (@eHealthUOC) February 27, 2024
The fifth grants call is now open
The eHealth Center has opened a new grants call for 2024, which this year is extended to include the UOC's Faculty of Psychology and Education Sciences, in addition to the Faculty of Health Sciences and the Faculty of Computer Science, Multimedia and Telecommunications.
The centre will award three grants of €3,000, one for each faculty, and the De la Idea al Projecte call also includes advice for recipients to be given by the team at the eHealth Center lasting a year.