An international research team used AI to check how genetics influence the structure of the left ventricle.
The studypublished in Nature, was carried out by the University of Manchester in collaboration with the University of Leeds, the National Scientific and Technical Research Council in Argentina and IBM Research in California.
Researchers used unsupervised deep learning to investigate over 50,000 3D MRI images from the UK Biobank, providing a basis for analyzing specific areas of heart structure for genetic associations using genome-wide and transcriptome-wide association studies (GWAS and TWAS).
The goal was to learn the way the structure of the center is said to genetics, thereby opening up opportunities for research into the results of genetics on organ formation and structure.
This could advance research into types of genetically determined congenital heart defects.
Here is a transient description of the way it worked:
- Data collection and processing: The team first used the UK Biobank database and chosen over 50,000 three-dimensional MRI images of the center. These images provided the fundamental data for the evaluation of the structure and morphology of the left ventricle.
- Training unsupervised models: Researchers then used unsupervised deep learning models to learn structures from these images, which meant the models identified patterns and features in the information without prior labeling.
- Extracting geometric features: With the unsupervised models installed, the team then focused on extracting geometric features from images depicting the left ventricle derived from the cardiac MRI data.
- Genome-wide and transcriptome-wide association studies (GWAS and TWAS): Armed with the extracted features, researchers conducted comprehensive GWAS and TWAS. These analyzes allowed them to check the connection between genetics and the structure of the left ventricle.
- Results: 49 latest genetic locations were identified with strong association with cardiac morphology, one other 25 were moderately associated.
Professor Alejandro F. Frangi explained the study by saying, “This is an achievement that might once have gave the look of science fiction, but we show that it’s entirely possible to make use of AI to know the genetic underpinnings of the left ventricle by “We only take a look at three-dimensional images of the center.”
Write on the University of Manchester BlogFrangi discussed the restrictions of previous studies and the breakthroughs made possible by these newer methods: “Previous studies have only examined the association of traditional clinical phenotypes… However, this study used AI not only to delineate the center chambers using three-dimensional medical images at pace, but additionally to create latest ones “to disclose genetic loci related to various cardiovascular deep phenotypes.”
The results of the study provide insights into the genetic basis of cardiovascular health and open latest avenues for the event of targeted therapies and precision medicine.
As Professor Bryan Williams of the British Heart Foundation described: “This latest research shows the big power of massive data, linking genes to heart structure.” Machine learning has made this possible by changing the way in which we do big Process, analyze and gain insights from data to reply the largest questions in cardiovascular research.”
AI models have previously been used to create detailed 3D maps of organs, including the human brain, for instance within the EU Large-scale Human Brain Project (HBP).
This goes a step further in linking genetics to organ structure and provides a deeper understanding of the morphology of the center and its genetic drivers.
This article was originally published at dailyai.com
