He mounted the lens so that its focal plane was on the scanner element, and sealed the thing to keep light out. That's pretty much it. It's a clever hack, notwithstanding the 4-5 second "exposure" time.
This kind of setup would actually be useful in practice for things like rare book scanning, where you can't flatten the pages for fear of damaging the object.
Size. Scanner is 8"x10" or so. The actual DPI of the scanner is lower than a traditional camera CCD, but there is so much more space. Your exposure time is how long it takes the scanner to traverse a "page".
Large format camera scanning backs have been around for quite a while, which is what this is. Big difference is that there is far less noise in the commercial versions.