iHEALTH researchers present advances in AI-powered medical imaging for Parkinson's, cancer and lung diseases
The Stream 4 meeting of iHEALTH — Population-based phenotyping for cardiovascular and liver disease — brought together researchers, clinicians and students at the UC San Joaquín Campus around three projects that show how artificial intelligence and advanced imaging can transform the diagnosis and monitoring of chronic diseases in Chile.
iHEALTH Principal Investigator and Stream 4 leader Cecilia Besa opened the session with an overview of the group's objectives and progress: to build a unique Chilean database —with both clinical and non-clinical images— that enables the development of AI-derived imaging biomarkers for chronic diseases, validating in our own population the tools developed across the institute's other Streams. A central challenge for the team is addressing model bias when facing demographically imbalanced cohorts.
The first presentation was given by Pamela González, who discussed her research on characterizing the brain changes that follow treatment with high-intensity focused ultrasound (HIFU) in Parkinson's disease. Using multimodal magnetic resonance imaging and brain-derived plasma analytes, her work seeks to understand what happens in brain networks beyond the clinical improvement already known.
As the researcher noted, "we know that HIFU already helps patients clinically, but we still don't understand what is happening in the brain network as a whole. Mapping how the focal lesion reorganizes those networks over time will allow us, in the near future, to anticipate outcomes and better select patients, turning HIFU into a more predictable and personalized therapy." González also stressed the local relevance of the topic: Chile is one of the countries with the highest projected prevalence of Parkinson's in the region by 2050.
The session also featured Universidad de Valparaíso doctoral student Esthefanía Astargo, who presented VIZCACHA (Visualization in Cancer With Assisted Chat), a web platform based on explainable AI and a conversational interface to support oncological diagnosis, with an initial focus on brain cancer using magnetic resonance imaging. The project aims to ease the workload of radiologists —in Chile there are only 6.6 radiologists per 100,000 inhabitants— without replacing existing hospital systems or clinical decision-making.
"The idea is for VIZCACHA to work as a support layer on top of the work the radiologist already does. The goal is to reduce the administrative burden and streamline their work, always keeping the clinical decision in their hands," Astargo emphasized.
The platform integrates three modules: a conversational interface to query reports and medical histories in natural language, an explainable brain tumor segmentation model, and a collaborative network to request second opinions among specialists.
The session closed with Carolina Pacheco, the Stream's new postdoctoral researcher, who presented the flagship project on functional lung magnetic resonance imaging for the assessment of chronic respiratory diseases such as asthma and COPD —conditions that affect more than 500 million people worldwide and are often diagnosed late. The team is developing low-field MRI techniques capable of capturing information with both spatial and temporal resolution, something that neither spirometry nor CT scanning can achieve on their own.
"This is a very open project: it hadn't been done before, largely because there were no techniques that truly allowed us to obtain spatial and temporal results at the same time. That's why we are very excited about this direction," said Carolina Pacheco, who acknowledged the earlier work of Carlos Valle, the project's previous postdoctoral researcher, and the team of collaborators led by Cecilia Besa and Marcelo Andia.
The meeting included a Q&A space after each presentation, with comments from clinicians and researchers who contributed new ideas to each line of work, reaffirming the collaborative and interdisciplinary character that defines iHEALTH's Streams.
