Purchase this article with an account.
Lucia M. Vaina, Martin A. Giese; Biological Motion: why some motion impaired stroke patients “can” while others “can't” recognize it? A computational explanation.. Journal of Vision 2002;2(7):332. doi: 10.1167/2.7.332.
Download citation file:
© ARVO (1962-2015); The Authors (2016-present)
Purpose: In its simplest way, Biological motion (BM) is defined by point-lights attached to the major joints and head of a human walker. To investigate the neuroanatomical areas necessary for BM processing we selectively “damage” a biologically inspired computational model of BM recognition to simulate the performance of stroke patients on this task.
Methods: We developed a hierarchically organized biologically driven computational model including both the ventral and dorsal pathway whose neurons exhibit a gradual increase in the complexity of the features processed. Neurons at the recognition level of the hierarchy are laterally connected, giving rise to recurrent network dynamics that can efficiently associate information over time. The model predicts quantitatively differences between the neural activation in different cortical areas. Driven by neuroanatomical and neurophyisiological considerations we used the model to perform simulations of (a) psychophysical results from15 stroke patients impaired on BM and (b) verify our results in fMRI (functional magnetic resonance imaging) studies of patients and normal subjects.
Results In the model, similar to the reports from neurological and fMRI studies, we found that the areas STP (superior polysensory temporal area) and area KO (“kinetic object” area) are necessary for recognition, as are either area MT+ (middle temporal area) in the dorsal pathway or the posterior and anterior IT (inferior temporal area) in the ventral pathway. fMRI results from our group and those of others support the functional-anatomical architecture of the model.
Conclusion: Impaired recognition BM demonstrated by patients with lesions in different cortical areas can be reliably accounted for by selectively “damaging” the network. We suggest that our model provides a biologically plausible functional architecture for biological motion recognition.
This PDF is available to Subscribers Only