Digital technologies for Healthcare

ITA has extensive experience in the field of data analysis, machine learning and deep learning applied both in the field of medicine and natural language processing, with several papers in research and a lot of successfully real application of machine learning, AI and NLP in production. Within the field of medicine, it has extensive experience working on several projects related to colorectal cancer -in the analysis and study of optimal decision criteria (risk factors) for screening programmes-, and gastric cancer, which aim to help detect cancers early. It also has experience in the development of diagnostic prediction models and decision-making support tools, in addition to omics and the study of biomarker combinations that optimise predictive capacity. It’s also important to highlight the research line related to liver transplant and the impact of comorbidity and other risk factors on patient survival after such a procedure. In fact, the analysis of comorbidities as a risk factor for cancers, transplant survival, etc., has been the motivation for several medical projects in which ITA has participated. The effect of comorbidities in the treatments and the evolution of chronic-kidney-disease patients is also being analysed in a current project, throughout the collection and analysis of data at national level. Within the NLP field, ITA has participated in several projects developing different NLP tasks such as the classification of text into categories, summarisation, named entity recognition or sentiment analysis, among others. NLP is also used in the generation of hybrid models that may help the healthcare professionals to predict a high probability of diagnosing certain types of cancer. Finally, we should mention our research line in image analysis, which is being currently applied in projects regarding sarcopenia and pose analysis. Technologies For the generation and development of the tasks/objectives, ITA applies and is expert in cutting-edge DeepLearning algorithms: long-short term memory neural networks (LSTM), bi-LSTM, convolutional neuronal networks (CNN), autoencoder models, transformers, Bert model, GPT models but normally our approaches usually mix Machine learning with ontologies or knowledge graphs. In addition to numerical and text data analysis, ITA also has expertise in image analysis. With all this knowledge, ITA always tries to combine in its projects different deep neural network architectures and hybrid approaches incorporating information from different data sources (numerical, text, image...) in order to improve the performance and accuracy of the model. Our group has extensive experience collaborating with institutions, companies, and SMEs in the healthcare sector to develop AI models and solutions for the prevention, diagnosis, prognosis, analysis of changes in patient behaviour, adherence, and personalized treatment of various diseases, based on patient data. We have solid experience in handling multimodal data, integrating numerical data (such as clinical-demographic, omics, lifestyle, medication use, sensor data, among others), textual data (using techniques for extracting and structuring medical record information, for example), and images, utilizing hybrid approaches to improve model performance. Additionally, we adopt a comprehensive ethical AI methodology, with an explainable and transparent AI approach that allows clinicians to trust the decision support tools developed. We also work on the research and innovation of conversational assistants, which can provide more efficient and personalized care, focusing on human-machine interaction and cognitive research.