I. Litvan et al., WHAT CAN ARTIFICIAL NEURAL NETWORKS TEACH US ABOUT NEURODEGENERATIVE DISORDERS WITH EXTRAPYRAMIDAL FEATURES, Brain, 119, 1996, pp. 831-839
Artificial neural networks (ANNs), computer paradigms that can learn,
excel in pattern recognition tasks such as disease diagnosis. Artifici
al neural networks operate in two different learning modes: supervised
, in which a known diagnostic outcome is presented to the ANN, and uns
upervised, in which the diagnostic outcome is not presented. A supervi
sed learning ANN could emulate human expert diagnostic performance and
identify relevant predictive markers in the diagnostic task, while an
unsupervised learning ANN could suggest reasonable alternative diagno
stic classification criteria. In the present study, we used ANN method
ology to try to overcome the neuropathological difficulties in differe
ntiating the subtypes of progressive supranuclear palsy (PSP), and in
differentiating PSP from postencephalitic parkinsonism (PEP) and corti
cobasal degeneration, or Pick's disease from corticobasal degeneration
. First, we applied supervised learning ANN to classify 62 cases of th
ese disorders and to identify diagnostic markers that distinguish them
. In a second experiment, we used unsupervised learning ANN to investi
gate possible alternative nosological classifications. Artificial neur
al networks input data for each case consisted of values representing
histological features, including neurofibrillary tangles, neuronal los
s and gliosis found in multiple brain sampling areas. The supervised l
earning ANN achieved excellent accuracy in classifying PSP but had dif
ficulty classifying the other disorders. This method identified a few
features that might help to differentiate PEP, supported currently pro
posed criteria for Pick's disease, corticobasal degeneration and typic
al PSP, but detected no features to characterize the atypical subtype
of PSP. In general, unsupervised learning ANN supported the present no
sological classification for PSP, PEP, Pick's disease and corticobasal
degeneration, although it overlapped some groups. Artificial neural n
etworks methodology appears promising for studying neurodegenerative d
isorders.