Artificial Intelligence applied to Healthcare
Description
This line of work is framed in the area of Bioinformatics and aims to provide novel solutions to problems related to precision medicine and offer help in the investigation of the mechanisms inherent to diseases. In this way, the application of AI techniques, more specifically Machine Learning (ML), and existing computing capabilities to the field of health will make it possible to generate support systems for physicians. This would facilitate their work and improve the quality of the services provided, resulting in a better quality of life for patients. Thus, the aim is to apply to healthcare techniques that have proven their validity and effectiveness in other fields, such as biometrics, handwritten text recognition, and automatic translation.
The digitization of healthcare data, with initiatives at local (hospital, clinic, etc.), national or international levels, generates an ever-increasing amount of genetic data, medical images, clinical, environmental, molecular data, etc. Machine learning feeds on this profusion of data by discovering patterns within it. Therefore, it is logical to want to apply ML techniques to the healthcare domain, and the set of applications that AI enables in healthcare is extensive. It explains why it is a line of R&D with significant development in the last decade. Of all the existing possibilities, PRAIA works mainly in treating and processing data through Machine Learning in general. The line is developed through close collaboration with different medical teams. The medical team is the one that guides the investigation of a specific disease using specific data, while PRAIA investigates ML techniques and applies them to provide solutions to the problems. The different diseases analyzed can be considered as use cases to implement a set of predictive services as generic as possible, simultaneously being able to respond to the needs of each disease studied and being easily extensible to other pathologies.
The fact of having the anonymized history of health data places the R&D line in the field of supervised machine learning. In other words, the actual target value (ground-truth) to be predicted is available. Healthcare data analysis trends are to have few observations (tens, hundreds) versus many variables (hundreds, thousands, and more). This situation is a real challenge for the following reasons:
- Inferring decision boundaries is a real challenge in high dimensionality spaces since they quickly tend to be very sparse and require many observations to draw reliable conclusions.
- The supervised problem is possibly incomplete: i.e., it is unknown to what extent we have patients who have the disease because of the variables included in the analysis or for another reason.
In short, this line of R&D aims to achieve a better understanding of the mechanisms underlying diseases and a more patient-centered medicine where treatments are tailored to the needs of each patient. All this is based on close collaboration with different medical teams and oriented towards developing functionalities or tools to support them in their work.