[its_so_on]

It clearly says, "the prediction is just categorical and is not an ‘image’ itself."

From the article:

>Neural decoding is performed using pattern recognition programs, also known as “decoders,” that learn statistical relationships between several pre-specified visual images and the corresponding brain activity patterns. Although this approach can predict which stimulus is presented among a set of pre-specified categories (e.g., horizontal or vertical grating), the prediction is just categorical and is not an ‘image’ itself.

In other words, this similar to saying, "Reconstruction of password using CPU activity" and then being able to tell which of a pre-specified list of passwords was used, based on clocking the CPU.

This is not exactly reconstruction.