What size net gives valid generalization?
Neural Computation
Learning internal representations by error propagation
Parallel distributed processing: explorations in the microstructure of cognition, vol. 1
Neural Networks for Pattern Recognition
Neural Networks for Pattern Recognition
| Hi-index | 0.00 |
The human brainstem is a highly complex structure where even small lesions can give rise to a variety of symptoms and signs. Localizing the area of dysfunction within the brainstem is often a difficult task.To make localization easier, we have developed a neural net system, which uses 72 clinical and neurophysiological data inputs and displays it (using 5268 voxels) on a three-dimensional model of the human brainstem. The net was trained by means of a back-propagation algorithm, over a pool of 580 example-cases. Assessed on 200 test-cases, the net correctly localized 83.6% of the target voxels; furthermore the net correctly localized the lesion in 31/37 patients. Because our computer-assisted method provides a reliable and quantitative localization of brainstem areas of dysfunction and can be used as a 3D interactive functional atlas, we expect that it will prove useful as a diagnostic tool for assessing focal brainstem lesions.