The medical images are a set of techniques and processes that serve to visualize the human body and diagnose diseases. In a short time, ct scans, X-ray imaging and ultrasound scanning in 2 dimensions (i.e., width and height) are read out in 3 dimensions. This advance in medicine will be possible according to say researchers at the MIT (Massachusetts Institute of Technology), they have created an artificial intelligence algorithm that allows you to retrieve the third dimension of the images with a single picture, that is to say, with a flat object. “The model could be used to recreate videos of images blurred by motion, or to create new types of cameras that capture the movement of a person,” says Guha Balakrishnan, a researcher of the computer science and Artificial Intelligence from MIT.
The algorithm is trained to retrieve information. “The visual data have a dimension that is lost completely, and it is the time or the space, and that is what this algorithm rescues,” says Balakrishnan. Although more evidence is needed until the algorithm is applied in the field of health, the scientists believe that this model could be used to convert medical images in 2D scans body in 3D. A very useful system in the developing countries for their low cost once perfected the algorithm. “You could do an x-ray and put it in our algorithm and to see all the information that was lost,” says Balakrishnan.
the researchers say that the three-dimensional world is only in our imagination or with imagination and 3D glasses. However, the reality is more complex. “When we have an object and we remove one dimension, everything becomes flat, in two dimensions. Imagine a pyramid viewed from above. The image is converted into four triangles planes”, explains Javier Melenchón, professor of computer science, multimedia and telecommunication Open University of Catalonia (UOC).
Another type of images that can be retrieved are the blurry photos, that is to say, when the shutter has captured a time and the object has moved. Is a projection through time. “Let’s imagine that we recover the function Live of iPhones (when there is a few tenths of a second of movement in a picture). With that service is projected to the time dimension. Is the height in pixels, the width of the pixels and projected the time,” notes Melenchón. In that case, it becomes an image in a video, it retrieves a variable, something similar to what you will do the scientists at MIT.
“visual data have a dimension that is lost completely, and it is the time or the space, and that is what this algorithm rescues”
The digital cameras that captured shots of long exposure added photons over a period of time in each pixel. When they capture the movement of an object along the time, the camera takes the average value of the pixels of the motion capture. Then, apply those average values of heights and widths corresponding to a still image, and create striped fuzzy on the trajectory of the object. “When we calculate some variations in the intensity of the pixels, we re-create in a theoretical way the movement,” explains Amy Zhao, engineering, computer science and electrical and researcher at the MIT. This algorithm recovers the original model it has been moved. “Would your projection, that is to say, the full expression for the object that has lost a dimension and, therefore, could be to make reconstructions of walls, tissues or even whole organs,” says Melenchón.
The researchers have been able to implement the algorithm after the system to recognize millions of couples of projection and the object that has caused the projection. From there the system follows the rules of projection to the three-dimensional object. This type of algorithms known as neural networks because they are inspired by the functioning of neurons, that is to say, automatically. “These algorithms deduce the internal rules once they’ve learned, not memorized,” says Melenchón.
