| OCR Text |
Show Recognition of Objects in Non- Canonical Views: A Functional MRI Study Kyla P. Terhune, MD, Grant T. Liu, MD, Edward J. Modestino, MPhil, Atsushi Mild, MD, PhD, Kevin N. Sheth, MD, Chia- Shang J. Liu, BA, Gabrielle R. Bonhomme, MD, and John C. Haselgrove, PhD Background: The neural correlate of object recognition in ( 3- 5). Patients with Alzheimer disease have been shown non- canonical views is uncertain, but there is evidence for to have more difficulty with this task as well ( 6). This in-involvement of neural pathways, possibly separate from formation suggests that processing objects in non- canonical those used for object recognition in canonical views. orientations uses a pathway either separate from, or in Methods: Boxcar functional MRI ( fMRI) techniques were addition to, traditional object recognition. The exact nature used to detect neural activity while eight normal subjects of these separate neural pathways has not been clarified were instructed to identify digital photographs of objects in ( 1,7). non- canonical and canonical orientations. In normal subjects, bilateral occipital cortical pro- Results: The right angular gyrus, the left inferior temporal cesses have been implicated in initial stages of object rec-gyrus, and the right cerebellum showed significant fMRI ognition ( 8- 10) and object shape processing ( 11). However, activity during non- canonical as opposed to canonical the right posterior cortex has been implicated clinico-viewing. pathologically in impaired recognition of objects in non- Conclusions: Subjects recognizing objects in non- canon- canonical views ( 3- 5). ical orientations engage in a process separate from, or in One cognitive process possibly involved in the recog-addition to, the process used in recognizing objects in canoni- nition of objects in non- canonical orientation is mental cal orientations. rotation. Functional MRI ( fMRI) studies have shown a wide . . . . _ , , , , -„„_ „ „_„ „_„. variety of correlates implicated in the mental rotation of ( J Neuro- Ophthalmol 2005; 25: 273- 279) , . . , . / 10 1^ j , , , , abstract objects ( 12- 15) and alphanumencal characters ( 16,17). These studies have reported activation of a variety of areas, including those within the parietal, frontal, and Objects appear in various orientations in the natural occipital lobes. The results of one recent fMRI study ( 18), environment. However, because of shape, gravity, and in which abstract computer- generated figures were used, the viewer's position, they are often seen in only a limited suggested that mental rotation and recognition of objects in number of orientations. Such standard orientations can be different views used different neural pathways, described as canonical, whereas unusual orientations can be Distinct from mental rotation is a three- dimensional described as non- canonical. Healthy subjects have been model ( 19) in which subjects are believed to store images shown to take longer to process and identify objects in non- based on geometry and volume. Marr ( 19) suggests that canonical positions than in canonical ones ( 1,2). When each physical object has minor and major axes, and these compared with controls, patients with right posterior cere- axes are important in deriving the geometry of an object and bral hemisphere lesions have demonstrated difficulties in then matching it to a subject's stored mental image. Under identifying common objects in non- canonical positions these circumstances, non- canonical perception is proposed to be more difficult if the major axis of the object is obscured. The study of Lange et al ( 20), however, provided Functional MRI Research Unit, Children's Hospital of Philadelphia, . , ... ,. , , , . . . , ~ ., , , , • „ , , TV • • r- AT n ,,, , , TT evidence that axes ot symmetry and elongation play rela- Philadelphia, Pennsylvania, and Division ot Neuro- Ophthalmology, Uni- . . . .. versity of Pennsylvania School of Medicine, Philadelphia, Pennsylvania. tivefy minor roles in object identification. Address correspondence to Grant T. Liu, MD, Division of Neuro- Warrington and James ( 21) observed that the angle of Ophthalmology, Department of Neurology, Hospital of the University of view and recognition were not always related, leading Pennsylvania, 3400 Spruce Street, Philadelphia, PA 19104; E- mail: gliu@ tQ ft m o d d b a s e d Q n ^ d i s t i n c t i v e k a t w e s 0 f a n o b j e c t. mail. med. upenn. edu T^- •• •• t- . t- • . • .- .1 o ^ A • ^ u *. r- r • I AT T i n + Distinctive features refer to unique properties of the Supported m part by the Clinical Neuroscience Irack Program at i r r the University of Pennsylvania School of Medicine ( grants NIH# Contours of the object, and in this model, the Spatial 1- R25- MH- 58818- 03 for KNS and KPT). relationship of one feature to another help identify the orientation and identification of the object. With further ( mean, 26 years). Seven were right- handed, and one was rotation of an object away from a canonical view, more of left- handed. these features would be obstructed. In related models, The subjects were paid to participate and received Hummel and Biederman ( 22) proposed that recognition complimentary copies of neuro- anatomical images of their of objects in different views depends on a binding of brain. Informed consent was obtained from all subjects analyses of individual parts, whereas Biederman and about the nature and consequences of the study. The proto- Gerhardstein ( 23) suggested that this process was possible col and study were approved by the Children's Hospital of when all views of an object contained the same " geon" Philadelphia Institutional Review Board, structural elements such as straight edges and curved sur- All subjects were free of abnormal neurologic or faces. Alternatively, Edelman and Duvdevani- Bar ( 24) theo- ophthalmic histories. Binocular best- corrected visual acuity rized that objects can be recognized by a relatively small was at least 20/ 25, and each subject had normal confron-number of reference shapes. ration visual fields. When required, each subject's manifest It may be that components of each of the above refraction or present prescription was used with a non-models are used in general object recognition. Davidoff and metallic lens set in a plastic frame ( adapted from a Titmus Warrington ( 5) reported a patient who exhibited significant stereo test; Gulden Ophthalmics, Abington, PA). Full difficulties in recognizing parts of an object but could still spherical and cylindrical correction was provided when recognize the same object in whole form in a canonical necessary. orientation. This observation suggests that the recognition The subjects' heads were secured firmly with foam of objects in canonical positions is not based on recognizing padding within the quadrature head coil to discourage individual features but on a broader form. Distinctive fea- motion. Volunteers were instructed to keep their heads still tares may be more important in the recognition of objects in at all times and their eyelids open during the periods of a non- canonical orientation. visual stimulation. A mirror was placed above the opening There are many explanations as to why it would be of the head coil and angled at 45 degrees so the subjects more efficient to store minimal canonical images and eval- could see a ground- glass screen ( Resonance Technologies, uate accordingly. First, the process requires minimal mem- Van Nuys, CA) placed at their feet. Once the subject was ory space. One need only store a common view or a few positioned in the MRI bore, dark material was placed on the common views and then use cortical processing to obtain sides of the bore opening to block the subject's peripheral alternative views. Second, the opportunity does not exist vision so that only the ground- glass screen was in view. to view every object encountered from all possible orientations. The ability to process objects at non- canonical orientations allows for the use of previous visual knowl- Stimuli edge at a much faster pace when presented with new A digital camera ( Nikon Coolpix 990; Nikon, stimuli. Melville, NY) was used to photograph common objects in Using fMRI paradigms and digital photographs of non- canonical ( 98 photographs) or canonical ( 97 photographs) common objects, this study aims to identify neural path- positions. Non- canonical views were denned as unusual ways involved in non- canonical identification. Photographs ones- usually foreshortened views. All objects were photo-are more realistic than computer- generated images and draw- graphed in black and white without flash when possible. The ings because they include cues such as depth and shadow. images were then converted to PICT files using Adobe Photo- Because the recognition of objects in non- canonical ori- shop ( Fig. 1). MacStim ( David Darby, Carlton, Australia), a entations is a natural behavior, it is important that the task Macintosh- based program, was used to animate the pre-be as similar to the natural environment as possible. There sentation of the photographs, which were projected onto the have been positron emission tomography ( PET) investi- ground- glass screen using a PLUS U2- 1080 video projector gations using drawings of common objects ( 1,2) and one ( resolution 1024 X 758). fMRI study using cropped photographs ( 25). No functional Black and white photography was chosen as opposed imaging studies have used photographs of objects in a to color photography, which might provide specific cues for natural environment. recognition. Tanaka and Presnell ( 26) have labeled objects low color diagnostic ( LCD) or high color diagnostic ( HCD). Objects classified as LCD were determined to be objects for which color was not a major identifier, whereas objects classified as HCD were determined to be objects Subjects where color was a major identifier ( such as a banana). We studied eight volunteers, consisting of five Using their findings, objects for which color was a major men and three women, ranging in age from 22 to 36 years determining factor were avoided. Boxcar Paradigm Siemens automatic shimming routine that used first and A total of eight epochs, four alternating pairs of epochs second order gradients. A " slice prescription procedure" containing objects either canonically or non- canonically was performed. First, a coronal scout image was obtained oriented were presented. Each epoch lasted 39.97 seconds, and oblique axial images perpendicular to the midline of and within each epoch, the duration of the presentation of this coronal image were prescribed. Subsequently, sagittal each image was self- paced by the subject via a fiberoptic, images perpendicular to the midline of the oblique axial push- button apparatus ( Current Design, Philadelphia, PA). images were taken. Finally, the 28 oblique axial planes The apparatus contained four buttons arranged in a covering the entire brain were acquired for the anatomic diamond- shaped pattern. After viewing a photograph of and functional images. They were positioned parallel to the one object, subjects were instructed to identify it silently by anterior commissure- posterior commissure line, name. They acknowledged identification by pressing the Tl- weighted neuroanatomical images were then right button and lack of identification by pressing the left obtained with a time to recovery ( TR) = 800 ms and a time button. With either choice, the program then advanced to to echo ( TE) = 15 ms. Twenty- eight axial slices, each 5 mm the next image. To determine if the subject correctly rec- thick, field of view ( FOV) 240 X 240 mm, and matrix 256 X ognized the object, the task was repeated outside of the 256 were acquired. T2- weighted echo- planar images were scanner, where the subject confirmed aloud the objects' acquired in identical planes as the Tl images. Twenty- eight names in the presence of an investigator. axial slices, with TR = 3.97 seconds, TE = 29 ms, flip Accuracy and average reaction times for each task angle = 90 degrees, 5 mm thick, FOV 240 mm X 240 mm, were calculated. In the object identification tasks, a response and matrix of 64 X 64 ( voxel size, 3.75 X 3.75 X 5cu/ mm) was scored as incorrect if 1) subjects determined during the were obtained. The boxcar experiment lasted 5 minutes and experiment that they could not identify the object, or 2) 35 seconds. A neuroradiologist reviewed the neuroanatom-subjects incorrectly identified the object when the task was ical images of each subject to exclude any abnormalities, repeated. Reaction times were calculated omitting those objects that subjects could not identify because reaction Post- Processine times were considerably longer for these images, and the ^ ^ were a n a l y z e d Qn a S u n S p A R C w o r k s t a t i on end point was not object identification but a random time ( Sun M i c r o s y s t e m s S u n S y s t e m > S a n t a c l a r a ; C A ) . T h e fcst arbitrarily determined by the subjects. Activation that oc- flye scans o f e a c h f ^ ^ e x p e r i m e n t w e r e discarded curred during the processing of non- identified images was fo e l i m i n a t e m a g n e t i c saturation effects. SPM99 ( Wellcome included within the functional analysis. Department of Cognitive Neurology, London, United Kingdom) was used to perform slice- timing correction, Scanning and Image Acquisition realignment, spatial normalization, spatial smoothing, and Imaging was performed using a 1.5 Tesla Siemens statistical analysis. Vision Magneton MRI Scanner system ( Siemens AG, Statistical slice- timing correction was performed to Munich, Germany). The magnet was shimmed using a account for the delay in the timing of data acquisition at different slices ( 27). To correct for motion, functional converted to Talairach ( 29) coordinates using a nonlinear images of each subject were then realigned using SPM99 transformation ( for Z > 0, X' = 0.9900X, Y' = 0.9688Y + to the first image by a six- parameter ( three translations and 0.0460Z, Z' = - 0.0485Y + 0.9189Z; for Z < 0, X' = three rotations) rigid body transformation. The images were 0.9900X, Y' = 0.9688Y + 0.0420Z, Z' = - 0.0485Y + then spatially normalized using SPM99 into the anatomical 0.8390Z). The Brodmann's Area ( BA) of these areas of space denned by the Montreal Neurological Institute ( MNI). activation were determined using the Talairach coordinates. This was accomplished by minimizing the sum- of- squares difference between the functional images and the SPM99 EPI template using an affine and nonlinear transformation. The voxel size was consistently maintained at 3.75 X 3.75 X 5 mm3. RESULTS Technically acceptable imaging studies and behav- Statistical Maps ioral responses were acquired for all subjects. The eight Spatial smoothing was performed with an 8 X 8 X subjects viewed an average of 19.9 non- canonical images 10 mm width at half maximum Gaussian kernel. Statistical and 23.0 canonical images per epoch ( 39.97 seconds), parametric maps, based upon the general linear model, were Mean reaction times for correctly identified objects were computed at each voxel. High- pass filter, low- pass filter, 1.707 seconds for non- canonical images and 1.542 seconds and global normalization were used. The reference wave- for canonical images ( t7 = 1.562; P = 0.162). Subjects form was modified by a hemodynamic response function. correctly identified a mean 85.8% of the non- canonical Activation during the non- canonical portions was con- photographs and 95.9% of the canonical photographs ( t7 = trasted with that during the canonical portions. Statistically 9.341; P < 0.001). An incorrect identification included significant areas across subjects were determined using both those that the subject could not identify during the test random effects analysis ( 28), which involved a first- level itself and photographs incorrectly identified by the subject analysis consisting of statistical parametric maps generated after the scan. using the general linear model and random field theory, When group data ( n = 8) were compiled using followed by the application of one sample t test to the SPM99, and non- canonical and canonical activation were individual activation maps as the second- level analysis. contrasted, three clusters of activation achieved statistical Areas of activation were determined in the following significance ( Table 1; Fig. 2). The first cluster was located manner: First, significant voxels were identified ( threshold in the right angular gyrus, extending into the superior P < 0.001, uncorrected at voxel level), then significant occipital gyrus and precuneus ( Brodmann areas 39/ 19; P< clusters among these voxels were identified ( threshold 0.001). The second cluster incorporated portions of the left P < 0.05, corrected at cluster level). These clusters were inferior temporal gyrus, the left middle occipital gyrus, characterized by the most significant voxel, but local and the left middle temporal gyrus ( Brodmann areas 37/ 19; maxima more than 8.0 mm apart from one another within P < 0.001). The third cluster was located primarily in the the cluster were also identified. MNI coordinates were then right cerebellum ( P - 0.014) ( Table 1). DISCUSSION bisected. Although spatial attention was believed to be re- Our experiments identified regions in the brain more quired to analyze the bisection of the line relative to the rest active while subjects recognized objects in non- canonical of the field additional visual spatial functions were believed views than while they identified objects in canonical views. to be used. These areas were BA 39/ 19 in the right angular gyrus Why were these areas more active while subjects region ( parietal lobe), BA 37/ 19 in the left inferior temporal identified objects in non- canonical views than in canonical gyrus region ( temporal lobe), and the right cerebellum. ones? Of the areas found to be active, BA 37/ 19 in the left The cortical regions identified in our experiment all inferior temporal gyrus region is most often implicated in have established roles in object recognition, visuospatial object recognition. One explanation for this increased acti-processing, or both. This is especially the case for the vation during non- canonical recognition is that an alternate temporal and parietal areas. For instance, the role of the left process in non- canonical recognition is used that actually occipito- temporal cortex in recognition of common objects relies more on object recognition pathways than does ca-has been shown clinically by McCarthy and Warrington ( 30), nonical recognition. Another explanation could be that the who detailed the case of an individual with visual asso- task of identifying objects in non- canonical orientations is ciative agnosia whose language, spatial, visual, and per- simply more difficult, leading to greater levels of activa-ceptual abilities were preserved but they had impairments tion. However, the average times that elapsed before sub-in object recognition. MRI scans showed an infarction in jects recorded responses in non- canonical and canonical the territory of the left posterior cerebral artery. Addition- conditions were not significantly different ( P = 0.162). This ally, Sergent et al ( 31) used PET to show that cerebral blood suggests that the difficulty of the two tasks was not as crit-flow in normal adults increases in this region when objects ical as the process itself. are recognized. Posterior cortical regions in both hemi- Activation of the other areas in the parietal lobe and spheres, but particularly in the ventral occipito- temporal cerebellum may be the result of recruitment of specialized regions, have been shown to be active in functional imaging visual spatial pathways, as discussed above, for recognition experiments studying object recognition in normal subjects of objects in non- canonical views. These areas may not be ( 10,32). active during recognition of common objects in more fa- There is also evidence to support the role of the miliar views ( 1,7,23,24). Clinical studies have demonstrated cerebellum in visual spatial processing, in addition to its the existence of separate pathways, particularly in the pos-more important one in motor coordination ( 33). For in- terior right hemisphere, which may mediate non- canonical stance, patients with cerebellar lesions have been observed object recognition. For example, when patients with right with deficits in visual spatial tasks ( 34). Furthermore, in posterior cerebral lesions were compared with patients with a fMRI study, Fink et al ( 35) showed neural activity in the left posterior cerebral lesions of comparable size and right posterior parietal cortex, the vermis, and the left cer- position, those with right- hemisphere lesions showed sig-ebellum when subjects performed a task involving visual nificant impairment when viewing objects in non- canonical spatial perception. Using the Landmark test, subjects deter- views ( 3). Vaina ( 36) presented a patient who was impaired mined whether transected horizontal lines were properly in non- canonical recognition, presumably as the result of bilateral lesions, right larger than left, involving the used in this study may limit the generalizability of the temporal- parietal- occipital junction. In Davidoff and results. Warrington's paper ( 5), a subject had a large infarct in Warrington and James ( 39) suggested that there may the temporo- parieto- occipital region of the right hemisphere. be an optional pathway used in object recognition when the Also, Landis et al ( 4) presented the case of a patient who standard, more direct route fails to provide recognition. In developed, in addition to other visual defects, an agnosia identifying cortical regions more active during recognition for real objects seen in non- canonical views. Autopsy of objects in non- canonical views than those in canonical revealed an occipitotemporal infarct in the territory of the views, our study supports this hypothesis, right posterior cerebral artery. Other functional imaging studies have attempted to localize the areas responsible for the process of non- Acknowledgments canonical identification. Using PET imaging, Kosslyn et al The a u t h o r s w o u l d l i k e t o t h a n k Dr- J iU H u n t e r for ( 1) found a variety of clusters, among them the left middle reviewing the Tl- weighted anatomical images and Drs. Martha temporal region ( BA 37) and the right angular gyrus ( BA Farah a n d Murray Grossman for their helpful suggestions. 39/ 19), thus complementing our results. Also reported were seven other clusters that reached significance ( P < 0.05). Apart from the mode of imaging, our object recognition REFERENCES study differed from that of Kosslyn et al ( 1) in other ways. . , , + i i • 1- Kosslyn SM, Alpert NM, Thompson WL, et al. Identifying objects Irrst, many extraneous thought processes took place in seen from different viewpoints: a PET investigation. Srazn 1994; 117: study of Kosslyn et al ( 1). Subjects were hearing stimuli, 1055- 71. seeing Stimuli, determining a match or lack of match, and 2- K o s s lyn SM> Thompson WL, Alpert NM. Neural systems shared by .. , , „ , , visual imagery and visual perception: a positron emission tomog-responding to the match. One would expect to see more raphy study Neumimage i997; 6: 320- 34. areas of activation than in our simpler paradigm. Also, 3. Warrington EK, Taylor AM. The contribution of the right parietal Kosslyn et al ( 1) used line drawings of objects rather than lobe t0 obJect recognition. Cortex 1973; 9: 152- 64. , , 4. Landis T, Regard M, Bliestle A, et al. Prosopagnosia and agnosia for photographs. noncanonical views: an autopsied case. Brain 1988; 111: 1287- 9. The Study most similar to OUrs is that of Sugio et al 5. Davidoff J, Warrington EK. The bare bones of object recognition: ( 25). In their fMRI experiment, subjects Were presented implications from a case of object recognition impairment. Neuro-psychologia 1999- 37- 279- 92 with cropped images of objects in non- canonical and canoni- 6 Caterini F> Della ' Sala s> Spinnler H> et al object recognition wA Cal positions. They were asked to identify them silently, SO object orientation in Alzheimer's disease. Neuropsychology 2002; 16: it is not certain whether the subjects were completing the 146- 55. . rrrl . . . . . . . 7. Burgund ED, Marsolek CJ. Viewpoint- invariant and viewpoint-task or not. The authors also compared activation during dependent object recognition in dissociable neural subsystems. non- canonical and canonical conditions within predeter- Psychon Bull Rev 2000; 7: 480- 9. mined anatomical regions of interest. They found greater 8- Malach R> RePPas JB> Benson RR> et al- object- related activity . . . . . ... . . revealed by functional magnetic resonance imaging in human occipital activation in the non- canonical viewing conditions than in cortex Proc Natl Acad Sci USA m5.92- M35- 9. the canonical viewing conditions for the premotor areas 9. Grill- Spector K, Kushnir T, Hendler T, et al. A sequence of object-bilaterally and the left superior parietal lobule. processing stages revealed by fMRI in the human occipital lobe. Hum ^ J .. . r J. / - Li ^ . . Bmin MaPP 1998; 6: 316- 28. One criticism ot our study might be that too many 10 Farah MJ A guirre GK. Imaging visual recognition: PET and fMRI extraneous tasks were required, because subjects Were asked studies of the functional anatomy of human visual recognition. to imagine the word mentally and then push a button to Trends Cogn Set 1999; 3: 179- 86. _ . , . _ . rT- . .... 11. Kraut M, Hart J Jr, Soher BJ, et al. Object shape processing in the confirm identification. The main concern was to avoid priming visual system evaluated using ^ t ^ MRL Neurology i997; 48: the subjects while still ensuring that the subject was focused 1416- 20. on the task and performing the exercise required. By 12- Cohen MS> Kosslyn SM, Breiter HC, et al. changes in cortical . . . , . . . .. activity during mental rotation: a mapping study using functional presenting only a photograph to subjects, priming was elrmi- MRI Bmin i996; ii9: 89- ioo. nated, allowing them to rely on their own mental images of 13. Richter W, Ugurbil K, Georgopoulos A, et al. Time- resolved fMRI the Objects, just as they would in a natural environment. of mental rotation. Neuroreport 1997; 8: 3697- 702. TT . . . _ . 14. Richter W, Somoriai R, Summers R, et al. Motor area activity during However, it was important to receive an answer from the mental rotation studied by time. resolved singie- trial fMRI. j Cogn subjects that confirmed that they were awake and par- Neumsci 2000; 12: 310- 2. ticipating fully, because attention to the visual stimulus has 15- Jordan K> Heinze HJ> Lutz K> et al- Cortical activations during , . . . . . . .-_-_. the mental rotation of different visual objects. Neuroimage 2001; 13: been shown to enhance cortical activation ( 37,38). Any 143- 52 extraneous motion, such as the movement of hands or 16. Alivisatos B, Petrides M. Functional activation of the human brain fingers to indicate answers, was present in both conditions durin § mental station. Neuropsychologic 1997; 35: 111- 8. . , .. 1 1 1 1 1 1 1 - 1 17. Harris IM, Egan GF, Sonkkila C, et al. Selective right parietal lobe in the boxcar paradigm and should have canceled in the activation during mental rotation: a parametric PET study Bmin contrast. Finally, the relatively small number of subjects 2000; 123: 65- 73. 18. Gauthier I, Hayward WG, Tarr MJ, et al. BOLD activity during 29. Talairach J, Tournoux P. Co- planar stereotaxic atlas of the human mental rotation and viewpoint- dependent object recognition. Neuron brain: an approach to medical cerebral imaging. Stuttgart: Thieme; 1988. 2002; 34: 161- 71. 30. McCarthy RA, Warrington EK. Visual associative agnosia: a clinico- 19. Marr D. Vision: A Computational Investigation into the Human anatomical study of a single case. J Neurol Neurosurg Psychiatry Representation and Processing of Visual Information. San Francisco: 1986; 49: 1233- 40. W. H. Freeman; 1982: 302- 9. 31. Sergent J, Ohta S, MacDonald B. Functional neuroanatomy of face 20. Large ME, McMullen PA, Hamm JP The role of axes of elongation and object processing: a positron emission tomography study. Brain and symmetry in rotated object naming. Percept Psychophys 2003; 1992; 115: 15- 36. 65: 1- 19. 32. Hasson U, Harel M, Levy I, et al. Large- scale mirror- symmetry 21. Warrington EK, James M. Visual object recognition in patients with organization of human occipito- temporal object areas. Neuron 2003; right- hemisphere lesions: axes or features? Perception 1986; 15: 355- 37: 1027^ 1. 66. 33. Rapoport M, van Reekum R, Mayberg H. The role of the cerebellum 22. Hummel JE, Biederman I. Dynamic binding in a neural network for in cognition and behavior: a selective review. J Neuropsychiatry Clin shape recognition. Psychol Rev 1992; 99: 480- 517. Neurosci 2000; 12: 193- 8. 23. Biederman I, Gerhardstein PC. Recognizing depth- rotated objects: 34. Schmahmann JD, Sherman JC. The cerebellar cognitive affective evidence and conditions for three- dimensional viewpoint invariance. syndrome. Brain 1998; 121: 561- 79. J Exp Psychol Human Percept Perform 1993; 19: 1162- 8. 35. Fink GR, Marshall JC, Shah NJ, et al. Line bisection judgments 24. Edelman S, Duvdevani- Bar S. A model of visual recognition and implicate right parietal cortex and cerebellum as assessed by fMRI. categorization. Phil Trans R Soc lond B Biol Sci 1997; 352: 1191- Neurology 2000; 54: 1324- 31. 202. 36. Vaina LM. Functional segregation of color and motion processing in 25. Sugio T, Inui T, Matsuo K, et al. The role of the posterior parietal the human visual cortex: clinical evidence. Cereb Cortex \ 99A; A: cortex in human object recognition: a functional magnetic resonance 555- 72. imaging study. Neurosci lett 1999; 276: 45- 8. 37. Watanabe T, Sasaki Y, Miyauchi S, et al. Attention- regulated activity 26. Tanaka JW, Presnell LM. Color diagnosticity in object recognition. in human primary visual cortex. JNeurophysiol 1998; 79: 2218- 21. Percept Psychophys 1999; 61: 1140- 53. 38. Somers DC, Dale AM, Seiffert AE, et al. Functional MRI reveals 27. Aguirre GK, Zarahn E, D'Esposito M. The variability of human spatially specific attentional modulation in human primary visual BOLD hemodynamic responses. Neuroimage 1998; 8: 360- 9. cortex. Proc Natl Acad Sci USA 1999; 96: 1663- 8. 28. Friston KJ, Holmes AP, Worsley KJ. How many subjects constitute 39. Warrington EK, James M. Visual apperceptive agnosia: a clinico-a study? Neuroimage 1999; 10: 1- 5. anatomical study of three cases. Cortex 1988; 24: 13- 32. |
