iHEALTH held first Principal Topics Seminar

With the virtual and face-to-face participation of researchers and students from the iHEALTH Millennium Institute, the first Symposium on Main Topics was held on April 26 at the Universidad Católica de Chile.

iHEALTH aims to develop innovative methods that integrate medical imaging, engineering and artificial intelligence (AI) to improve image-based medical care with the long-term goal of making it a more affordable and accessible alternative for all Chileans.

“To fulfill its purpose, iHEALTH has the support of researchers from different disciplines and who work on issues that point towards the same from a wide variety of areas and methods”, explained its Director Claudia Prieto.

The symposium was organized with the aim of presenting the lines of research that are currently being carried out so that all members of the Institute could learn what the work of the rest of the team consists of and thus contribute to the meeting and to the synergy of everyone who makes up the Millennium Institute.

The invetigator who started the presentations was Prof. Pablo Irarrazaval who presented an introduction to the physics of magnetic resonance imaging (MRI, for its acronym in English). When asked about the main challenges in his line of research, Dr. Irarrazaval replied that they can be summed up in two: “On the one hand, we have that the speed at which magnetic resonance images are acquired is slow and we need that they be faster to, in this way, lower costs for both health systems and patients. On the other hand, there is a whole line of research that deals with obtaining quantitative measurements in the images, this in order to deliver useful information to reach more precise diagnoses”, he explained.

Then it was the turn of our Director, Prof. Claudia Prieto, who spoke about the reconstruction of medical images and commented to the audience that there are multiple approaches and methods to make more assertive medical diagnoses. Reconstruction of medical images is an inverse problem because we want to generate an image of the interior of the human body, starting from some measured physical information. The role of artificial intelligence (AI) has proven to be a great contribution to these types of problems and in medical imaging in general, for example, to speed up the acquisition of images and facilitate their analysis to generate a medical diagnosis.

Prof. Domingo Mery then presented an introductory talk on AI. As he explained, AI includes methods such as machine learning, or automatic or machine learning, and deep learning. “Deep learning is very useful for image classification, however for this model to work it requires a lot of images that must be previously classified by human beings. This classification becomes increasingly complex if the images are not “easily” recognizable by everyone, as is the case with medical images that require “expert eyes”. Finally, the professor emphasized the importance of acting ethically when working with artificial intelligence and, above all, with the use of sensitive or personal information, such as medical images.

Prof. Jocelyn Dunstan and Denis Parra shared a talk on the role of Natural Language Processing and Generation in artificial intelligence with applications in medicine. In Medicine, explained Prof. Jocelyn, there are some classic problems related to the detection of relevant information within a document or text prepared by a doctor, with the correct coding of diseases in the international classifications that they exist for this (CIES) and, also in the field of ethics, with the anonymization of medical records. Regarding the work with iHEALTH, the purpose is to use a type of analysis called multimodal and that includes integrating medical images with written text, which is done through neural networks.

For his part, Prof. Rodrigo Salas, commented that one of the biggest problems he seeks to solve with his lines of research in iHEALTHs has as a starting point the lack of experts who can interpret the information contained in medical images. Mainly, he explained that his work seeks to develop new artificial intelligence models that allow explaining, both to medical professionals and to patients, how diagnoses are reached. “Most AI models are black boxes that need to be explained to doctors and patients, that is, to explain how the diagnosis of having or not having cancer is reached, for example, explaining what the reasons are,” he added.

The person in charge of closing the session was Prof. María Fernández who spoke about the role of AI in physiological signals and sensors, which facilitates not only interpretation, diagnosis and prognosis; but also treatment and rehabilitation in health. “One of the main benefits of AI is that it can combine information, and not only in terms of medical images, but with a large amount of physiological data such as blood pressure, temperature, etc.” pointed out.

The day ended with a round of questions where the attendees were able to share their queries and insights to reach agreements and understand the work carried out at iHEALTH. “We come from different areas, therefore these types of meetings, which we will continue to seek in the future, are very beneficial to achieve a conceptual base that allows us to get closer and communicate in a better way and thus propose collaborative works”, concluded Prof. Claudia Prieto.