Houston, we ve a tumour
what can dock the space shuttle and spot a brain tumour? Not a medically-qualified astronaut but a neural networking computer algorithm designed by engineers at the us National Aeronautics and Space Administration ( nasa ). The system has become so adept at analysing the large amount of data needed to dock a spacecraft that it is also being employed by doctors to analyse tumours.
Recognising patterns is neural networks forte. And they do it pretty well with a small number of parameters, but go beyond 10 or so inputs and the neural nets get confused and become hard to train.
For humans, by contrast, this sort of task comes quite naturally. For instance, a surgeon looking at a chunk of tissue might subconsciously analyse its colour, consistency and shape, and conclude that it is a malignant growth. All this can happen in a split second, without the surgeon being aware of the whole process.
Now a neural computer that handles many pieces of data at once has been designed by a team led by Robert Mah, an engineer at nasa 's Ames Research Center in California. "It's a neural net where lots of parameters are analysed in real time,' explains Mah. "It's looking at 25 or more; it's pretty hard to do that.'
Mah's neural net can take the data from a surgical probe