The latest advances in techniques for artificial intelligence they are becoming more and more related to the world of medicine and health. Although difficult tasks have been automated, current techniques require large databases to learn from.
In many cases, it is difficult to compile databases with thousands of images of each thing that the method to be developed needs to know, either because the images are difficult to obtain or because an expert is needed to invest a lot of time in labeling them.[banner-DFP_1]
In this aspect, a work carried out by researchers from the Visilab research group, from the Higher Technical School of Industrial Engineering of the University of Castilla-La Mancha (UCLM) and published in the journal “Computer Methods and Programs in Biomedicine”, through of applications based on computer vision and image processing for diagnosis, has proposed a method that will improve biomedicine.
They have created a new system to generate new images by mixing two of those already available in a small database
To solve the problem of the lack of data in medicine or biology by creating artificial images, they have created a new system to generate new images by mixing two of those already available in a small database.
The work written by the responsible researchers Noelia Vállez, Gloria Bueno and Óscar Déniz, members of the Visilab group and professors of the Higher Technical School of Industrial Engineering; and Saúl Blanco Lanza, from the Institute of Environment, Natural Resources and Biodiversity of the University of León, shows a developed method that is inspired by the life cycle followed by diatoms, microscopic algae found in rivers and seas.[banner-DFP_4]
The researchers give an example to understand this new discovery: “it is easy for a person to learn what a chair is by seeing two or three, but for a computer learn the same thing, you need to see photos of thousands of different chairs and someone to “teach” you and teach you that these are chairs.
When augmented databases are used with the proposed method, based on image morphing and registration, the results obtained by artificial intelligence techniques show a clear improvement in accuracy.
“It would be like taking two images that we know contain chairs and mixing them to obtain a chair of an intermediate size and shape,” says researcher Vállez.
The work shows that when augmented databases are used with the proposed method, based on image morphing and registration, the results obtained by artificial intelligence techniques show a clear improvement in accuracy.
In this way, the improvement was measured in different databases related to medicine and biology to demonstrate its applicability in different fields and problems.
Because we all need health… ConSalud.es