29 July 2021
AperTO - Archivio Istituzionale Open Access dell'Università di Torino
Original Citation:
Spatiotemporal profiles of visual processing with and without primary visual cortex
Published version:
DOI:10.1016/j.neuroimage.2012.07.058
Terms of use:
Open Access
(Article begins on next page)
Anyone can freely access the full text of works made available as "Open Access". Works made available under a
Creative Commons license can be used according to the terms and conditions of said license. Use of all other works
requires consent of the right holder (author or publisher) if not exempted from copyright protection by the applicable law.
Availability:
This is the author's manuscript
This full text was downloaded from iris - AperTO: https://iris.unito.it/
iris - AperTO
University of Turin’s Institutional Research Information System and Open Access Institutional Repository
This Accepted Author Manuscript (AAM) is copyrighted and published by Elsevier. It is
posted here by agreement between Elsevier and the University of Turin. Changes resulting
from the publishing process - such as editing, corrections, structural formatting, and other
quality control mechanisms - may not be reflected in this version of the text. The definitive
version of the text was subsequently published in NEUROIMAGE, 63 (3), 2012,
10.1016/j.neuroimage.2012.07.058.
You may download, copy and otherwise use the AAM for non-commercial purposes
provided that your license is limited by the following restrictions:
(1) You may use this AAM for non-commercial purposes only under the terms of the
CC-BY-NC-ND license.
(2) The integrity of the work and identification of the author, copyright owner, and
publisher must be preserved in any copy.
(3) You must attribute this AAM in the following format: Creative Commons BY-NC-ND
license (http://creativecommons.org/licenses/by-nc-nd/4.0/deed.en),
10.1016/j.neuroimage.2012.07.058
The publisher's version is available at:
http://linkinghub.elsevier.com/retrieve/pii/S1053811912007884
When citing, please refer to the published version.
Link to this full text:
Spatiotemporal profiles of visual processing w ith and w ithout primary visual cortex
Andreas A. loannides
, Vahe Poghosyan , Lichan Liu , George A. Saridis , M arco Tam ietto
Marc Op de Beeck , X avier De Tiège , Lawrence W eiskran tz , Beatrice de Geldei
The spatiotem poral profiles o f visual processing are norm ally distributed in tw o tem poral phases, each last ing about 100 ms. W ith in each phase, cortical processing begins in V I and traverses the visual cortical hier archy. H ow ever, the causal role o f V I in starting each o f these tw o phases is unknown. Here w e used m agnetoenceph alography to study the spatiotem poral profiles o f visual processing and the causal contribu tion o f V I in three neurologically intact participants and in a rare patient (G Y ) w ith unilateral destruction o f V I , in w h o m residual visual functions m ed iated by the extra-geniculostriate pathw ays have been reported. In healthy subjects, visual processing in the first 200 ms post-stimulus onset p roceeded in the tw o usual phases. N orm ally p erceived stimuli in the left h em ifield o f GY elicited a spatiotem poral p rofile in the intact right hem isphere that closely m atched that o f healthy subjects. H ow ever, stimuli presen ted in the cortically blind h em ifield produced no detectable response during the first phase o f processing, indicating that the re sponses in extrastriate visual areas during this phase are d eterm in ed by the fe ed forw a rd progression o f ac tivity initiated in V I. The first responses occurred during the second processing phase, in the ipsilesional h igh -level visual areas. The activity then spread fo rw a rd tow a rd h igh er-level areas and backward tow ard lo w e r-le v e l areas. H ow ever, in contrast to responses in the intact hemisphere, the back-propagated activity in the early visual cortex did not exh ibit the classic retin otopic organ ization and did n ot have w ell-d e fin e d response peaks.
Introduction
A substantial portion o f conscious vision is m ediated b y the geniculostriate visual pathway, w hich relays visual inform ation from the retina to the prim ary visual cortex ( V I ) and then to the extrastriate cortices. The cortical processing in this pathw ay begins in V I at about 4 0 -5 0 ms post-stimulus onset and proceeds rapidly through lo w - and m id-level retinotopic extrastriate areas (V 2 -V 5 ) (Grill-Spector and Malach, 2004; Lamme and Roelfsema, 2000; Poghosyan and loannides, 2007). W ith in the first 100 ms after stimulus exposure activity spreads
beyond the retinotopic areas proceeding tow ard high-level, largely non-retinotopic visual areas (loannides and Poghosyan, 2012; Liu e t al., 2009; M eeren e t al., 2008; Okazaki e t al., 2008). The second phase o f ac tivity through the same areas occurs in the next 100 ms interval (loannides and Poghosyan, 2012; M eeren e t al., 2008; Sugase e t al., 1999). W ith in each phase cortical processing begins in V I and traverses the visual cortical hierarchy (loannides and Poghosyan, 2012).
Substantial visual processing is also m ediated b y the extra- geniculostriate pathways that bypass V I and project directly to the extrastriate cortex. In fact, patients w ith cortical blindness follow in g dam age to V I can retain som e residual visual capability in the absence o f awareness ( “blindsight" phenom enon) (P op p el e t al., 1973; Weiskrantz, 2009b; W eiskrantz e t al., 1974). Hemianopic patient GY, w h o sustained selective early dam age to his left V I can effectively discriminate stimuli presented in his (righ t) blind hem ifield (C o w e y and Stoerig, 2004; de Gelder et al., 1999; Schurger e t al., 2006; Tam ietto e t al., 2009, 2010; W eiskrantz et al., 1995; Zeki and Ffytche, 1998). Neuroim aging studies on patient GY have found that his intact right occipital lobe and the intact portions o f his left (lesion ed ) occipital lobe show conventional retinotopic organization (Baseler et al., 1999). H ow ever, w h en stimuli are presented
in the blind iiem ifield, activity is restricted to dorsal and/or ventral extrastriate areas and shows abnormal retinotopic organization (Barbur e t al., 1993; Goebel et al., 2001; Sahraie e t al., 1997; Tam ietto e t al., 2010; Zeki and Ffytche, 1998).
Notably, h o w e ve r, no study has id en tified the precise spatiotem - poral sequence o f visual cortical responses fo llo w in g presentation o f stimulus in the cortically blind hem ifield, o r has d irec tly contrasted these spatiotem p oral profiles w ith those elicited b y the sam e stimuli in th e intact hem ifield. Furtherm ore, it is not y e t k n ow n w h e n and w h e re the first such cortical response appears, o r h o w it spreads th ereafter alon g the ad jacent cortical areas, and w hat, i f any, is the causal contribu tion o f V I to this activity.
In the present study w e used m a gn etoen cep h a lograp h y (M E G ) t o g e th e r w ith a distribu ted source m od el (loa n n id es e t al., 1990; T aylor e t al., 1999) to estim ate w ith m illisecond tim e accuracy the spatio- tem p ora l p rop erties o f neural a c tiv ity e licited b y checkerboard- pattern stim uli placed in d ifferen t p ortions o f the visual field. W e first d ocu m en ted the k ey spatiotem p oral features o f visual stimulus processing in th ree norm al subjects, substantially e xten d in g the anal ysis o f data rep orted e a rlier (P og h osya n and loannides, 2 00 7 ). W e th en described the neural processing in the intact and d am aged hem isp h ere o f G Y in response to the sam e stim uli and in referen ce to that o f the th ree con trol subjects.
Methods
Subjects
P a tien t GY
G Y is a 5 6 -year-old m an w ith righ t h em ian opia and “b lind sigh t" fo llo w in g selective d am age to his le ft V I suffered at the age o f 7, as the result o f a traum atic brain injury. GY's visual system has b een p re v iou sly tested in b eh avioral and psych oph ysiological exp erim ents, as w e ll as w ith fM R l and d iffu sion-ten sor im a gin g m ethods, see (B aseler e t al., 1999; G oeb el e t al., 2001; T am ietto and de Gelder, 2010; T am ietto e t al., 2 01 0 ). T he procedu re used to map GY's visual field in the current study is described in S upplem entary m ethods.
The MEG experim ent w ith G Y was conducted at the MEG unit o f the ULB-Hopital Erasme, Brussels. The resident ethical com m ittee approved the study, which was perform ed in accordance w ith the ethical standards
laid d ow n in the 1964 Declaration o f Helsinki. G Y gave inform ed consent to participate in the study.
Controls
Three healthy right-handed m ale subjects aged 2 6 -2 8 years, partic ipated in the MEG experim ent. A ll had norm al vision. For each subject, the exp e rim e n t w as repeated on three d ifferen t days. A ll experim ents w ith norm al subjects w e re conducted at RIKEN Brain Science Institute in W ako-shi, Japan. RlKEN's ethics com m ittee ap p roved the study, and all the subjects ga ve inform ed consent. In an earlier publication (P oghosyan and loannides, 2 00 7) w e reported the analysis o f the a ver aged data from a subset o f the runs from this exp e rim e n t w ith emphasis on the reproducibility o f the results across d ifferen t exp erim ental days.
S tim u li and experim ental design
P a tien t GY
GY was com fortably seated in a m agnetically shielded room (M SR). The stimuli w e re delivered via a DLP projector (M o d e l PT-D7700E, Panasonic, N e w Jersey, USA) located outside the M SR Images w e re back-projected on a screen inside the MSR v ia an optical periscope.
Nine different locations o f the visual field w e re stimulated, one at a tim e, using circular checkerboard patterns. The same parameters w e re used as for the control subjects for luminance and contrast levels (see Controls section), w ith slight m odifications in size and locations made to ensure that the stimuli presented in the right visual hem ifield fell en tirely w ith in the blind field o f GY. Specifically, each stimulus had a radius o f 4° and a check size o f 1 ° ( Fig. 1 A ) . N orm ally the stimuli w e re presented centered on the visual vertical meridian (3.6° above and b e lo w the fixa tion), horizontal m eridian (1 0° to the left and right o f the fixation) and in the visual field quadrants 10° horizontally and 3.6° vertically o ff the visu al meridians: on the upper vertical m eridian (U V M ), lo w e r vertical m e ridian (L V M ), left horizontal m eridian (L H M ) and right horizontal m erid ian (R H M ), and in th e u p p er rig h t (U R ), lo w e r rig h t (LR ), u p p er le ft (U L ) and lo w e r le ft (L L ) visu al field quadrants. T he central (n e a r -fo v e a l) stim ulus w a s cen te re d 1.5° to th e le ft o f th e fixa tion in o rd e r to b e e n tire ly w ith in th e intact p ortio n o f th e visu al field . The stim ulus in th e UR qu ad ran t w a s cen te re d s ligh tly clos e r (1 ° ) to the h orizon ta l m erid ian (a t 2.6° ab o ve th e m e rid ia n ) than th e h o m o lo gous stim ulus in th e UL quadrant. In this w o r k w e focus on the
B
H
V
' ' )
)
• * • A *... .V’ I
y \ ...:K
.‘ V...V. /Fig. 1. Schematic im age o f stimuli. A. Stimuli used in the experim ent w ith GY. Circular checkerboard patterns w ith radius o f 4° w e re presented one at a tim e on the vertical and
horizontal visual meridians, and in each quadrant o f the visual field. Norm ally the stimuli w ere placed along the vertical lines passing through the central fixation (visual vertical m eridian), and 10° to left and right o f it; and along the horizontal lines passing through the central fixation (visual horizontal m eridian), and 3.6° above and b elo w it. The stimulus at the fixation (fo v e a ) was centered 1.5° o ff the fixation into the left visual field so as to fall entirely in the intact portion o f the visual field. The stimulus in the upper right quadrant was centered at 10° to the right o f the fixation (as the normal stim uli) and at 2.6° above it. All stimuli had a check size o f 1 °. B. Stimuli used in the experiments w ith controls. Circular checkerboard patterns w ith radii o f 2.5° and 4° w ere presented one at a tim e on the vertical and horizontal visual meridians, and in each quadrant o f the visual field along its 45° diagonals in random order at 4° and 9° eccentricities, respectively. All stimuli had a check size o f 1°.
com p a rison b e tw e e n the responses to stim u li fallin g e n tire ly w ith in the b lind part o f th e rig h t visu al h em ifie ld and w ith in th e h o m o lo gous location s o f th e sighted le ft h em ifie ld . The results from the v e r tical m eridian stimulations are p rim arily used for delineation o f the early visual areas; w e w ill not consider further the data from the near-foveal stim ulation in the current manuscript.
In one run, each visual field location was stimulated 10 times. Five such runs w e re recorded, thus across the runs each visual field location was stimulated 50 times. Locations w e re stimulated in random order. Each stimulus was displayed for 500 ms, w ith an interstimulus interval random ized b etw een 400 and 600 ms. The subject was instructed to fix ate on a small red cross at the center o f the display throughout the run and to try not to make any eye m ovem ents. GY was questioned at the end o f each o f the 5 runs about his visual experience. In all occasions G Y reported no awareness w hatsoever o f visual stimuli presented in his blind hemifield.
Controls
T he subjects w e r e com fo rta b ly seated in a MSR. The stim uli w e re d eliv ered via a fib e r-op tic g o g g le system (A v o te c Inc., Jensen Beach, FL, USA), w h ic h subtended 3 0 ° x 2 3 ° o f visual angle.
Sixteen d iffe r e n t location s o f th e visu al field w e r e stim u lated, one at a tim e, using circu lar ch eckerb oa rd patterns, w it h th e check size o f 1 ° ( Fig. IB ). T he lu m inance o f th e dark and lig h t checks, as m easured w ith a p h o tom eter, w a s 5 and 90 cd/m^ re sp e ctive ly, thus the M ich elson con trast and m ean lu m inance o f th e checkerb oa rd pattern stim uli w e r e 89% and 47.5 cd/m^ re sp e ctive ly. Stim uli at tw o d iffe r e n t eccen tricities, 4° (p a ra fo v e a l) and 9° (p e rip h e r a l), w e r e p re sented cen te re d on th e visu al v e rtic a l m erid ian (u p p e r and lo w e r to th e fix a tio n ), h orizon ta l m erid ian (le ft and righ t to th e fixa tio n ) and in th e visu al field qu adrants a lon g th e 4 5° diagonals. P arafoveal and p erip h era l stim u li had radii o f 2.5° and 4° re sp e ctive ly. T he la rg er size o f th e stim ulus p resen ted at th e p erip h era l location s d id not c om p en sa te fu lly for the cortical m a gn ifica tion factor (R o v a m o and Virsu, 1979).
S tim u li at d iffe r e n t e cc e n tric itie s (i.e. at 4° and 9 °), and on and o f f (i.e. in th e q u a d ra n ts) visu al m e rid ian s w e r e p res e n ted in d iffe r e n t runs. H en ce in each run th e stim ulus w a s p res e n ted at fou r v is u al field location s, a to ta l o f 240 tim es - 60 to each location . Locations w e r e s tim u lated in ra n d om ord er. Each stim ulus w as d is p la y e d for 500 ms, w it h an in terstim u lu s in te rv a l ra n d o m ize d b e t w e e n 4 00 and 600 ms. T he su b ject w a s instru cted to fixa te on a sm all red cross at th e c e n te r o f th e d is p la y th ro u g h ou t th e run and to tr y n ot to m ake an y e y e m o v em e n ts .
M R ! data preprocessing and coregistration
Controls
The h igh-resolution anatomical MRls o f control subjects w e re taken w ith a 1.5-T Siemens M agn eton Sym phony scanner in Tokyo, Japan. The cortical reconstruction and volu m etric segm entation w e re perform ed using the same fully-autom ated Freesurfer analysis pipeline as was used for GY.
The co-registration o f MEG and M Rl data for each control subject w as p erform ed using a special m ethod that provides a co-registration accuracy o f about 1 m m (H ironaga and loannides, 2002). The details o f the p erform ed procedures can be found e lsew h ere (P oghosyan and loannides, 2007).
S u pporting fM R I experim ent
The supporting fMRI experim ent w ith control subjects was conducted using a Varian Unity Inova 4-T w h ole-b od y MRI system (Varian N M R In struments, Palo Alto, CA) equipped w ith a M agnex head gradient system (up to 3.4 gauss/cm w ith a rise tim e o f ~ 160 |js, M agnex Scientific Ltd., Abingdon, UK). During the recording, the subject's head was fixed using a bite bar. T w o pressure sensors w e re also placed around the subject's head to m onitor any head motion. The functional data w e re acquired using multishot T2*-w eighted echo-planar imaging pulse sequence (T R = 6 0 0 m s ; T E = 1 5 m s ; flip angle = 40°, sampling bandwidth = 102.4 kHz). M ore details o f the fMRI experim ent can be found elsew here (Cheng e t al., 2001).
The stimuli w e re delivered to the subject's eyes via a pair o f indepen d en tly controlled fiber-optic bundles (Silent Vision binocular glasses, A v otec Inc., Jensen Beach, FL, USA) subtending 3 0 °x 2 3 ° o f visual angle. To identify the borders b etw e en V I and V2 visual areas vertical meridian was stimulated using a vertically oriented b o w tie-shaped black/white checkerboard w ith contrast reversal at 7.5 Hz on a gray background.
M EG data acquisition and preprocessing
P a tien t GY
MEG signals for GY w e re collected at the MEG unit o f the ULB-Hopital Erasme, Brussels using the 306-channel (204 planar gradiom eters and 102 m agnetom eters) w hole-head V ectorview MEG system (Elekta Neurom ag Oy, Helsinki, Finland). The detailed characteristics o f the sys tem can be found elsew here (D e T iege e t al., 2008). The MEG signals w e re recorded w ith a passband o f 1 -3 3 0 Hz at a sampling rate o f 1000 Hz. Vertical and horizontal electrooculogram (FOG), and electrocar diogram (ECG) w e re recorded in synchrony w ith MEG signals. The posi tion o f the subject's head relative to MEG sensors was continuously m onitored b y feeding current to four head localization coils attached to the scalp.
P a tien t GY
A high-resolution anatomical M Rl o f GY was taken w ith a 3-T Siemens M agneton Allegra scanner in Erlangen, Germany. Cortical reconstruction and volum etric segm entation w e re perform ed using a fully automated analysis pipeline provided w ith the Freesurfer image analysis suite, w hich is docum ented and freely available for dow nload online (http:// surfer.nmr.mgh.harvard.edu/). G Ys lesion was manually traced on the high-resolution M Rl and was projected on to the cortical surface using a softw are utility d eveloped in-house (s ee Supplementary m ethods).
T he p osition s o f the MEG sensors w e r e c o -re g is te re d to the subject's an atom ical M Rl using th e fo llo w in g p rocedu re. B efore each e x p e rim e n t fou r head loca liza tion coils w e r e attached to the scalp o f GY. The entire surface o f his head, togeth er w ith the head localization coils, w as d igitized w ith respect to anatom ical fiducials (nasion, right and left tragus) using a 3D d igitizer (Polhem us 3Space/ Fastrak, USA). All the d igitized points w e re then fit to the M Rl head shape, provid in g the position o f the MEG sensors w ith respect to the subject's M Rl w ith an estim ated accuracy o f 3 mm.
Controls
MEG signals for norm al subjects w e r e c ollected at the Laboratory for Hum an Brain Dynam ics (1 9 9 8 -2 0 0 9 ) at RIKEN Brain Science Insti tute, Japan using a 151 g ra d io m e te r w h o le -h e a d O m ega system (CTF/ VSM M ed Tech Ltd., BC, Canada). In synchrony w ith the MEG signals, vertica l and horizontal FOG and ECG w e r e record ed at the sam e sam p ling rate o f 1250 Hz. In addition, e y e position d uring the w h o le run w as record ed w ith an e y e tracking system (A v o te c Inc., Jensen Beach, FL, U SA ) m ou n ted on th e goggles.
The subjects' head location w a s record ed at the b egin n in g and end o f each run, w h ich lasted about 4 min. A v e ra g e head m o v e m e n t o f each subject d uring a run w as around 1 -2 m m . Runs in w h ic h m o v e m en t exc e ed ed 4 m m w e r e repeated.
Offline, the MEG signals w e re con verted to a 3 rd-ord er synthetic gradient, low -pass filtered at 200 Hz, and resam pled at 625 Hz. Trials w h e re the eyes deviated m ore than 1° from the fixation cross, or w h e re the subject blinked, w e re identified using the eye tracker and FOG data, and w e re rem oved from further analysis.
Data analysis
A ll MEG and M R Id ata (fr o m G Y and the three control subjects) w e re analyzed in the same w a y at the Laboratory for Human Brain Dynamics at A A l Scientific Cultural Services Ltd., Cyprus. In d ependent com ponent analysis (Jahn e t al., 1999) in conjunction w ith ECG data w as used to re m o v e the cardiac artifacts. The cleaned MEG signals w e re then d ivid ed into 600 ms trials, from — 200 to 400 ms w ith respect to stimulus onset.
Tom ographic source analysis
M agnetic field tom ograp h y (M P T ) (loannides e t al., 1990; Taylor e t al., 1999) w as applied to single trial MEG signals to estim ate the three-dim ensional distribution o f current sources throughout the brain, in dep en d en tly for each snapshot (tim e poin t o f a single trial) o f data. M PT uses a non-linear algorith m for solving the biom agnetic inverse problem . It estim ates the prim ary current distribution j (^r'j inside the source space S w ith the follo w in g recursive expression:
r e s
(
1
)
w h e re are scalar expansion coefficients, w hich must be determ ined togeth er w ith j „ r ) in e ve ry iteration, r ) is the initial a prio ri prob ability w e ig h t function, w h ich in practice is determ ined independently for each experim ental arrangem ent (i.e. for each run and subject) through a training session w ith a know n set o f current dipoles spread throughout the source space, and are the lead field functions, w hich relate j ( ) to the m agnetic field measured at MEG sensor:
: - j ( r ) d r , k = { i , . . . , N )
(2)
By com b in in g Eqs. (1 ) and (2 ) in e v e r y iteration w e obtain a set o f equations:
iJ, = j ’ i k i r j ' j w o ( 7 ) [ ; „ ( ? ) d ? ,i = (1, ...,JV) (3 )
w h ere from Ail can be determ ined. Then b y substituting Ail in Eq. (1 ),
j n ( j ^ is found. To overcom e the ill-posedness o f the Eq. (3 ) in the soft
w are im plem entation o f MPT Tikhonov regularization is used, w ith the regularization parameter pre-determ ined using com puter simulations. In the current study and in all our earlier studies the current source distri bution, j ( ? ), has been estimated using one iteration ( j i ( 7 ) ), w ith a starting choice o f j o ( j ) = 1 • The detailed analysis o f lead field expan sions has show n that the standard form o f MPT has optim al stability and sensitivity for localized distributed sources (Taylor et al., 1999). In con trast to linear inverse methods, standard MPT relies less on the raw lead fields and makes m ore use o f the measured data. Specifically, it makes less assumptions than the linear methods, because in the standard MPT
only the direction o f the current source distribution is expressed b y a lin
ear expansion in lead fields, w h ile the linear methods presuppose that both direction and strength o f the current source distributions can be expressed in a single linear expansion in the lead fields.
Por each subject, five partially overlapping source spaces, covering the left and right hemispheres, and front, back and top o f the brain have been defined. MPT was used to estimate the current source distribution sepa rately in each source space using signals from the 90 MEG sensors best covering the source space under examination. The MPT estimates from all five source spaces w e re then com bined together, w ith a w eighting that reflects the sensitivity profile o f the inverse solution at each source space point, yielding a final source space that covers the entire brain.
P ost-to m ogra p h ic source analysis
Sets o f single trial M PT estim ates w e r e then used to com pu te a tim e-v a ry in g sign al-to-n oise ratio (S N R ) (Laskaris and loannides, 2001; Raz e t al., 1988) map for each stim ulated location o f the visual field, exp e rim e n ta l d a y and subject. The SNR w a s c om pu ted using the fo llo w in g form ulae:
A — N , p(N-l)
P
i-2 JVSNR = N P
w h ere X i{t,p ) is a p ms interval centered at t ms latency and extracted from the single trial. The SNR values for each brain v oxel w e re com puted using 20 ms w in d ow s (i.e. p = 2 0 ) m oving from — 190 to 390 ms (i.e. t = - 1 9 0 ... 390). Each SNR map was generated from a set o f 50 to 60 single trial MPT estimates and in all figures presented in this paper SNR values above 0.2 are shown. This SNR threshold was se lected so as to elim inate all pre-stimulus activity in control subjects. The m eth od ology was successfully applied in a num ber o f recent studies, always show ing peak activations in V I in close agreem ent w ith the know n retinotopic organization o f V I : Stimuli falling in one visual hem ifield (i.e. UR, RHM, LR, LL, LHM and UL) elicited activations in the contralateral brain hemisphere, w h ile stimulations o f the lo w e r and upper visual field locations (i.e. UVM, UR, LR, LVM, LL and UL) activated regions dorsal and ventral to calcarine sulcus respectively.
Regions o f interest and regiona l tim e courses
SNR maps w e re visually inspected tim eslice-by-tim eslice and brain re gions that exhibited distinct activations consistently across the three e x perim ental days w e re identified. Por each subject the centroids o f such activations w e re designated as centers o f spherical regions o f interest (R O l) w ith a radius o f 1 cm. The principle direction for each ROl, which is the dom inant orientation o f the current density vectors in that ROl was calculated using circular statistics (Pisher, 1993), an established fram ew ork for identification o f statistically significant distributions based on both magnitude and direction o f vectors (loannides et al., 2005; Poghosyan and loannides, 2007, 2008). Por control subjects and the unaffected visual hem ifield o f the patient GY, ROls w e re sought starting from the first suprathreshold activation on each SNR map till 100 ms after stimulus onset. Por three stimuli (UR, RHM and LR) that fell in the cortically blind hem ifield o f GY, ROls w e re defined from the first suprathreshold responses elicited b y them; for each o f these three stimuli the first response was identified at latencies longer than 130 ms and w ithin the high-level visual cortex. Three such ROls w e re defined in the ipsilesional left hemisphere. Three additional ROls w e re defined, one for each “unseen" stimulus, in the same (ipsilesional left) hem isphere from the first responses in the early visual cortex (area V2 or V3). N ext w e identified three ROls in each control subject corresponding to the three high-level ipsilesional ROls o f GY. Such ROls w e re d efined based on ana tomical and functional criteria. Anatom ically they had to be w ithin 1 cm from their respective patient ROls in the Talairach space (Talairach and Tournoux, 1988) and have similar localization w ith respect to sulcal/ gyral landmarks — the ROls w e re around m iddle occipital gyrus (M O G ), superior tem poral sulcus (STS) and m iddle tem poral gyrus (M TG ). Func tionally, their localization and latencies had to be reproducible across the three experim ental days.
Regional activation curves (RAC) w e re then generated for all V I ROls in intact hem ispheres and the three high-level ipsilesional ROls, for both controls and the patient. RAC describes the activation tim e course o f an ROl along its principal direction and w e re generated from single trial MPT estimates b y integrating, for each tim e point o f 1.6 ms, the projec tions o f the current density vectors along the principle direction in the ROl. To quantify the signal content in an ensem ble o f single trial RACs w e com puted their signal p ow er (SP) (Laskaris and loannides, 2001; Raz e t al., 1988) using a 10 ms m oving w in d ow . SP tim e courses w e re
generated separately for each ROl, stimulated visual field location and subject.
C om puter sim ulations
The data for G Y and control subjects w e re recorded at different sites using tw o different MEG systems. Our analysis m eth odology is robust enough to adequately deal w ith this limitation, as earlier studies have demonstrated (loannides e t al., 2002). W e nevertheless confirm here, using com puter simulations, the consistency o f the source localization across the tw o systems and the ability to identify and distinguish tw o con currently active nearby sources (in V I and V2, -0 .9 cm apart). The details o f com puter simulations are provided in the Supplementary methods.
Results
Analysis o f sim ulated data
The analysis o f simulated data show ed that the localizations obtained from G Y and our controls subjects can be directly com pared despite the differences in the MEG hardw are used to record the signals in each case, and that the e m p lo ye d m eth o d o lo gy is capable o f identifying concurrentìy active source in V I and V2 (s e e Supplem entary results and Supplem entary Figs. 1 and 2).
Source analysis in the in ta ct hemispheres
Controls
The first SNR activation (a b o ve the threshold o f 0.2) in both brain hem ispheres o f the norm al subjects was identified b etw een 40 and 70 ms post-stimulus, w ith its peak localized in V I. These peak activa tions w e re in good agreem ent w ith the know n retinotopic organization o f V I (H olm es, 1945; Horton and Hoyt, 1991): the activations w e re con tralateral to stimulus hem ifield, stimuli in the lo w e r and upper visual fields activated regions dorsal and ventral to calcarine sulcus respective ly, and activations in response to parafoveal stimulations w e re localized m ore posterior and closer to the occipital pole com pared to activations elicited b y peripheral stimulations, w hich w e re d eeper in the calcarine. The responses elicited b y stimuli presented along the visual meridians (i.e. UVM, RHM, LVM and LHM ) w e re localized w ith in 2 -3 m m o f the corresponding fMRI representations, w hich w e re obtained for normal subjects in an independent fMRI exp erim ent (Fig. 2A).
Following this initial response in V I the activity rapidly spread through the early visual cortex, ventrally tow ard occipitotem poral and dorsally tow ard posterior parietal brain regions. The results for one nor mal subject are show n in Fig. 3A (LL stimulation) and 3B (U L stimulation). The points w ith 50% o r higher probability for cytoarchitectonic areas V4 and V5 are also displayed on the inflated brains. The figure shows that the activity reached the end-level o f the classic retinotopic visual cortex, area V4, at - 8 0 ms post-stimulus. In agreem ent w ith earlier studies (Fylan e t al., 1997; Portin et al., 1999; Tzelepi e t al., 2001) activity in the early visual cortex was stronger in response to lo w e r than upper visual field stimulations.
P a tien t GY (in ta c t hem isphere)
The spatial and tem poral pattern o f visual responses in the intact hem isphere o f G Y was similar to that o f control subjects (using the same SNR threshold o f 0.2). Specifically, the first brain responses to left visual field stimulations w e re identified in the right V I at “ retinotopically correct" locations b etw e en 50 and 70 ms post-stimulus (Fig. 2 A right part). Thereafter, the activity rapidly traversed the visual cortical hierar chy (Figs. 3C and D). Again responses w e re stronger for lo w e r rather than upper visual field stimuli.
Regions o f interest in the in ta ct hemispheres
W e used the first 100 ms o f the post-stim ulus period to discern n o table SNR activation peaks in norm al subjects and in the intact h em i sphere o f GY. For each stim ulated location w e then d efined ROls centered at the corresponding peaks. In both brain hem ispheres o f nor mal subjects, the distribution o f the ROls on the flattened occipital c or te x follo w ed the classic visual field maps (Fig. 2 A ). In norm al subjects stimuli w e r e presented in the quadrants and on the m eridians o f the v i sual field at tw o d ifferen t eccentricities. The borders b etw e en early visu al areas could thus be estim ated b y connecting the representations o f stimuli placed at d ifferen t eccentricities on each m eridian on the flat tened occipital cortex using the fo llo w in g procedure. The ROls obtained from the first activations elicited by stimuli presented along the upper and lo w e r vertical meridians (U V M and LV M ) d efin ed resp ectively the ventral and dorsal borders b etw e en areas V I and V2. The ROls obtained from the n ext activations elicited b y the same stimuli (U V M and LVM ) d efin ed the borders o f V3 and V4. For the data analyzed here the V I - V2 b order w as identified in all hem ispheres but the V 3 -V 4 b order o n ly in som e hem ispheres. The ROls obtained from the first activations elicited b y stimuli presented along the le ft and right horizontal m e rid ians (LH M and R H M ) separated the ventral and dorsal portions o f V l . The ROls obtained from the next activations elicited b y the same (LH M and R H M ) stimuli d efined the borders b etw e en areas V 2 and V3 in the right and left brain hem ispheres respectively. This putative b o r d er allocation w as substantiated b y three in dependent results. First, stimuli p resented in visual field quadrants elicited activity in early visu al areas in -b etw een the corresponding representations o f vertical and horizontal meridians. The entire set o f ROls w as p erfectly consistent w ith the retinotopic maps obtained routinely w ith fMRI. Furthermore, the latencies and the sequence o f activations from w h ich the ROls w e re d efined w e re in accord w ith a feed forw a rd spread o f activation. Second, as can be seen in Fig. 2A, our V I - V 2 b order estim ates (w h ite lines) are in excellen t agreem en t w ith the independent fM RI estim ates (black lines). Finally, the relative alignm ent o f the retinotopic areas V I - V3 d efin ed from our MEG-based source analysis and the areas V 4 and V5 d efined from the probabilistic cytoarchitectonic maps (E ick hoff e t al., 2 00 5) is consistent w ith the know n o rganization o f visual cortex: V4 and V 5 fall nicely ju st outside the retinotopic areas V I- V 3 .
The range o f stim uli used for G Y w as m ore lim ited, but still, in the intact hem isp h ere the ap p roxim ate borders b e tw e e n e a rly visual areas can be d efin ed b y lines con n ecting the ROls for stim uli along the m eridians w ith the occipital p ole (rig h t part o f Fig. 2 8 ). These ROls w e r e d efin ed fro m the e arly responses at latencies sim ilar to the ones id en tified in norm al subjects. The estim ates for the V 1 -V 2 and v en tral V 2 -V 3 borders w e r e consistent w ith the results from the control subjects. Furtherm ore the first activations e licited by stim uli ju st o f f the horizon tal m erid ian fall n icely w ith in the e xp e cted cytoarch itecton ic area.
Tim e courses in the in ta ct hemispheres
To e xa m in e the tem p oral dynam ics o f individual ROls for d ifferen t stim ulated locations, w e produ ced RAC-based SP tim e courses from single trial MFT estim ates, for all V I ROls and the th ree “ ipsilesional" ROls, for norm al subjects and for the p atien t (s e e R egions o f interest and region al tim e courses section ). In norm al subjects, the V I tim e courses for the retin otop ica lly ap p ropriate stim uli w e r e h igh ly re p ro ducible across d iffe re n t e xp erim en tal days as alrea dy rep orted for qu adrant activations (P og h osya n and loannides, 2 00 7 ). Fig. 4 shows V I tim e courses for all stimuli; the curves from each o f the three exp er im ental days are superim posed dem onstrating the reproducibility o f responses. The figure shows also the separation into processing phases, d efined b y the tw o distinct responses, in the 5 0-1 00 and 100-200 ms latency ranges. These results (Fig. 4, see also Fig. 9 ) together w ith our re cent findings (loannides and Poghosyan, 2012) and other converging
Fig. 2. Distribution o f the ROIs on the flattened occipital cortex. The left hemisphere flattened patches are shown on the left and the right hemisphere patches on the right. Colored
shapes mark centers o f ROIs obtained in response to stimuli presented in different parts o f the visual field, as depicted on the upper right part o f the figure: Cyan and y e llo w shapes mark ROIs corresponding to stimuli presented at eccentricities o f 4° and 9° respectively. Triangles, rhombi and circles indicate ROIs for stimuli presented on the horizontal meridian, vertical meridian and in the quadrants respectively. Filled and em pty shapes indicate ROIs for stimuli presented in the upper and lo w e r visual fields respectively. The markings for estim ated visual area borders are indicated on the exam ple flattened patch on the upper right part o f the figure: W h ite lines indicate borders betw een early visual areas estim ated based on source analysis o f MEG data alone. Black lines indicate the borders betw een areas V I and V2 estim ated in independent flVIRI experiments. Putative V4 and V5 areas obtained from the above 50% cytoarchitectonic probabilistic maps are also indicated on the flattened patches. A, Distribution o f ROIs in a typical control subject. B. Distribution o f ROIs in GY. Crosses mark the locations o f the three h igh-level ROIs in the ipsilesional hemisphere. The black patches show the lesioned portion o f the left occipital cortex.
evidence (Liu e t aL, 2002; M eeren et aL, 2008; Sugase et aL, 1999) siiow tiiat in tiie first 200 ms after stimulus onset normal visual processing contains at least tw o phases, spanning roughly the first and second
100-ms post-stimulus intervals respectively. During each phase activity starts in V I and spreads thereafter in a feedforw ard direction through the hierarchy o f visual cortical areas. In accord w ith earlier studies
72 ms
Fig. 3. Spread o f activity from the initial response in V I as a function o f time. For each tim e instant the activations (S N R > 0 .2 ) are shown on the lateral, medial and ventral aspects o f
inflated cortical surfaces o f right hemispheres. Putative V4 and V5 areas obtained from the above 50% cytoarchitectonic probabilistic maps are also shown and are indicated on the first lateral and ventral view s o f the figure. A, B. Typical examples taken from one control subject for the stimuli presented in the LL (A ) and UL (B) quadrants at 9° eccentricity. C, D. Results taken from GY for the stimuli presented in the LL (C) and UL (D ) quadrants.
(Fylan e t al., 1997; Portin e t al., 1999; Tzelepi et al., 2001) and the e v i dence from our SNR maps, SP tim e courses w e re stronger in response to lo w e r than upper visual field stimuli.
Fig. 5 show s tim e courses for the patient's intact (rig h t) hem isphere V I ROIs, d efin ed separately for UVM , UL, LHM, LL and LVM stimuli (see Regions o f interest and regional tim e courses section). No noticeable V I activation w as seen for stimuli presented in the cortically blind right h em ifield (n o t show n). In all V I ROIs the strongest SNR w as found in re sponse to retinotopically corresponding visual field location (s h o w n in Fig. 5). In line w ith the results from norm al subjects, the visu ally evok ed activity in the intact hem isphere contained tw o distinct responses, at - 7 0 and 150 ms latencies, corresponding v e r y lik ely to the tw o-phase visual processing schem e observed in the norm al subjects. Further m ore, SP w as higher in response to stimuli presented in the lo w e r than upper visual field, especially during the first phase o f processing.
Results fro m the left (ipsilesiona l) hemisphere o f GY
W e studied n ext the a c tiv ity e licited b y stim uli p resented in GY's cortically blind righ t h em ifie ld (UR, RHM and LR). T he m ost striking d ifferen ce b e tw e e n the activations in the d am aged hem isp h ere c o m pared to the ones in th e intact h em isp h ere w as the absence o f a c tivity during the first phase o f processing (< 1 3 0 m s). Setting the sam e SNR threshold o f 0.2 as the on e e m p lo y e d for th e con trol subjects and for the intact hem isphere, the first activations in th e d am a ged h em i sphere e licited b y stim uli p resented in the patient's b lind h em ifield (rig h t h em ifield ; UR, RHM and LR) ap p eared b e tw e e n 140 and 160 ms post-stim ulus in the d am aged hem isp h ere (w h it e crosses in the left part o f Figs. 2B, and 6 ). These first responses in the d am aged hem isp h ere are show n in Fig. 6 superim posed on M Rl anatom ical
slices and inflated cortical surface. For ease o f referen ce the putative M T -I-/V5 accordin g to cytoarch itecton ic p rob a b ility maps (M a lik o v ic e t al., 2 0 0 7 ) is also displayed on the inflated maps. The th ree id en ti fied activations w e r e localized in the m id d le occipital gyrus (M O G , u p p er r o w in Fig. 6 ); around the conflu ence o f superior tem p ora l sul cus (STS), m id d le tem p oral gyrus (M T G ) and M OG (m id d le r o w in Fig. 6 ); and on the b ord er o f M TG and STS (lo w e r r o w in Fig. 6).
T he le ft part o f Fig. 28 show s the locations o f the peak activity (w h it e crosses) for each o f th e th ree areas o n the flat m ap o f the d a m aged hem isp h ere to g e th e r w ith the p utative V 4 and V 5 areas. The ab sence o f retin otop ic activations in e arly visual corte x did n ot a llo w us to d efin e the visual area b orders in the d am aged hem isphere. H o w e v er, the ju x tap osition in Fig. 2B o f the le ft hem isp h ere w ith the intact righ t hem isp h ere stron gly suggests that these first activations in the d am a ge hem isp h ere are in the h igh -lev e l visual cortex.
F ollow in g these first responses in the d am aged hem isp h ere ac tiv ity spread forw a rd to occip itotem p ora l corte x and backw ard to early visual corte x (Figs. 7A, B and C). The locations o f the first responses in the e arly visual corte x e licited b y stim uli p resented in the cortically b lind righ t h em ifie ld (UR, RHM and LR) are show n in Fig. 2B (y e llo w shapes in th e le ft p art). The responses to UR and LR stim uli lik ely fall w ith in area V2 o r V3, w h ile the first response to RHM stim uli is w ith in V 3 A o r V7. Alth ou gh the LR e licited response is in a brain region w ith a b o ve 50% p rob a b ility for V 4 based on cytoarch itecton ic probabilistic maps, its location relative to calcarine sulcus/lesion and the ju x ta p o sition o f le ft and righ t h em isp h ere flat maps indicate that the re sponse is in ven tral V2 or V3. H o w e ve r, in contrast to responses in the intact hem isp h ere (Figs. 7D, E and F) these late ‘feedb ack ’ activa tions d id n ot ad h ere to classic retin otop ic organization : responses to u p p er and lo w e r visual field stim uli w e r e loca lized in dorsal and
0.8
-100 O 100 200 Time (ms)
300
Fig. 4. SP tim e courses for V I ROls in a typical control subject. Tim e courses obtained
from three different experim ental days are superimposed (red - day 1, blue - day 2, green - day 3). Each figure displays tim e courses for one location o f the visual field (printed on its upper left part). The placem ent o f each figure corresponds roughly to the location in the visual field and proceeding in a clock-wise direction - they are UVM, UR, RHM, LR, LVM, LL, LHM and UL. The results fo r stimuli presented at 9° eccen tricity are shown. Each figure shows the tim e courses fo r each day for the ROl corre sponding to the stimulated location.
ven tral e arly visual corte x resp e ctive ly (s e e the y e llo w shapes in the le ft part o f Fig. 2B and zo o m e d insets in Figs. 7A and C). T h e y w e re also less o rga n ized in tim e (Fig. 8).
Fig. 9 shows SP tim e courses for the patient in the three high-level ipsilesional ROls (n o te the absence o f response during the first processing phase, < 1 30 ms). Both left and right visual hem ifield stimuli produced prom inent responses in these ROls, confirm ing that they are located b e yond the classic retinotopic cortex and in high-level visual areas. The re sponse latencies o f all three ROls for left (ipsilateral) hem ifield stimuli (retinotopically mapped to the intact hem isphere) w e re v e ry similar, starting around 120 ms post-stimulus and peaking at ~ 150 ms. The laten cies in response to right (contralateral) hem ifield stimuli (retinotopically m apped to the dam aged hem isphere) w e re slightly longer, stronger and m ore variable across stimulus locations.
Finally, w e identified three ROls in each norm al subject correspond ing to the GY'S three ipsilesional h igh -level ROls (s ee Regions o f interest and regional tim e courses section). Fig. 10 show s the SP tim e courses in these three ROls in the representative norm al subject. The responses in these three left h em isphere ROls for the norm al subjects and G Y can be contrasted in the con text o f tw o-ph ase visual processing. The responses in norm al subjects (Fig. 10) sh ow tw o clear visual processing phases: in all three ROls the responses in the first phase peak around o r just b efore 100 ms post-stimulus for the contralateral stimuli (rig h t visual field stim uli) and about 20 ms later for the ipsilateral stimuli. In the second phase, the peak activity is earlier in the left MTG (g re e n lines) at -1 5 0 ms post-stimulus, w h ile the activity in the other tw o ROls peaks at -2 0 0 ms. For the patient no evok ed activity w as found b efore 120 ms in the dam aged hem isphere. The latencies o f the first responses
UVM
Time (ms)
Fig. 5. SP tim e courses for V I ROls in the (righ t) intact hemisphere o f GY. Each figure
displays tim e course in the same form at as in Fig. 4, but only for the visual fields: UVM, LVM, LL, LHM and UL. As in Fig. 4, each figure shows the tim e course for the ROl corresponding to the stimulated location.
in these three ipsilesional ROls ap p roxim ate those o f the second-phase responses in the norm al subjects. The responses from the dam aged hem isphere o f G Y th erefore suggest that in the absence o f V I the entire first visual processing phase is missing. The responses in each o f these three high -level visual areas in the patient peaked roughly at the same latency (Fig. 9 ), sim ilar to the first-phase responses in normal subjects (Fig. 10). The latencies o f responses elicited b y stimuli in the cortically blind contralateral hem ifield w e re 1 0 -2 0 ms lon ger c o m pared to responses in the same areas elicited b y stimuli in the sighted ipsilateral hem ifield.
Discussion
Our results show ed that the visual processing in norm al subjects in the first 200 ms o f the post-stim ulus period p roceeded in tw o phases, spanning roughly the first and second 100-ms post-stim ulus intervals, respectively, in good agreem en t w ith oth er studies (loannides and Poghosyan, 2012; Liu e t al., 2002; M eeren e t al., 2008; Sugase et al., 1999). In each phase the stim ulus-evoked activity started in V I and quickly spread through the dorsal and ventral streams tow ard higher levels o f visual cortical hierarchy up to p ost-retinotopic extrastriate areas (loannides and Poghosyan, 2012).
T h e e a r ly visu a l p ro ce ss in g e lic ite d b y n o rm a lly p e rc e iv e d s tim uli p ro je c te d in th e le ft h e m ifie ld o f G Y p ro d u ce d a c tiv ity in his in ta ct (r ig h t ) o c c ip ita l lo b e w h ic h f o llo w e d th e s p a tio te m p o ra l p a ttern o b s e rv e d in h ea lth y subjects, in clu d in g th e can onical re tin o to p ic o rg a n iz a tio n (B a s e le r e t al., 1 9 9 9 ), re sp o n se tim in g and seq u en ce, and s e g m e n ta tio n in to t w o d is tin c t p ro ce ss in g phases. H o w e v e r , visu a l p ro ce ss in g in th e d a m a g e d ( l e f t ) h e m i s p h ere fo r u n seen r ig h t h e m ifie ld s tim u li w a s q u a lit a tiv e ly d iff e r ent. In fact, no c o rtic a l re sp o n se w a s e v id e n t at la ten c ies c o v e r in g
Upper right quadrant
Cyto area
MT+/V5
Left MOG BA19
Right liorizontal meridian
SNR from MEG
144 ms
Left MT+A/5
Lower right quadrant
Left MTG BA37
Fig. 6. The first activations in response to “unseen" stimuli in the blind hem ifield o f GY. The first activations (S N R > 0 .2 ) elicited by stimuli presented in the UR quadrant (upper row,
MOG BA19), on the RHM (m id dle row, M T + A ^ 5 ) and in the LR quadrant (lo w e r row, MTG BA37) are shown on the MRl and lateral aspect o f the inflated cortical surfaces o f the left hemisphere. Note that the first responses to all three “unseen” stimuli w e re in the ipsilesional hemisphere. Axial (le ft), sagittal (m id d le) and coronal (rig h t) MRl slices best covering the relevant activations are shown. The y e llo w contours on each MRl v ie w encompass the regions w ith S N R > 0.2. Activation latencies are given b elo w the sagittal view s. The black rectangles on the inflated cortical surfaces indicate the zoom ed areas shown in their lo w e r right p a rt Putative V5 area obtained from the 50% cytoarchitectonic probabilistic maps is also shown in the inflated maps and they are captured also on the first and second zoom ed view s.
the first processing phase (< 1 3 0 m s). This clearly indicates that the re sponses in extrastriate visual areas observed in the intact brain during this phase are determ ined b y the feed forw a rd progression o f activity initiated in V I. The first activations elicited b y stimuli in the cortically blind h em ifield occurred after 140 ms in the contralateral (ipsilesional) h igh -level extrastriate areas around IVIOG, IVITG and STS. F ollow in g this first response, the activity spread forw ard rapidly through tem poral and occipital cortices tow ard higher level areas and backward tow ard early retinotopic visual areas. The back-propagated activity reached earlier visual areas such as V2/3 in 1 0-2 0 ms. H ow ever, in contrast to re sponses in the intact hem isphere, this activity did not fo llo w the classic retinotopy. Specifically, stimuli presented in the lo w e r right quadrant o f visual field (LR ) elicited activity p redom inantìy in the ventral V2/3, w h ile the activity in response to stimuli presented in the upper right quadrant o f visual field (U R ) w as in dorsal V2/3. This spatial organiza tion o f responses does not agree w ith the predictions o f classic retinotopic m odel, according to w h ich LR and UR stimuli should p ro duce responses in dorsal and ventral portions o f the early visual cortex respectively. H ow ever, the back-propagated activity m ay fo llo w non-classic retin otop y resulted from the reorganization o f the visual c ortex fo llo w in g the lon g-term lesion o f V I (Rosa e t al., 2000).
T h e d e lin e a tio n o f e a rly visu al cortic a l areas o f n orm al subjects p ro v id e d h ere is c on s isten t w it h th e m aps o b ta in e d ro u tin e ly w ith fIVIRl. For each c on tro l subject, th e V I - V 2 b ord e rs (re p re s e n ta tio n s o f th e v e rtic a l m e r id ia n ), e stim a ted in th e c u rren t stu d y on th e basis o f sou rce analysis o f IVIFG d ata alone, w e r e in e x c e lle n t a g re e m e n t w it h th o s e ob ta in e d fro m an in d e p e n d e n t fIVIRl e x p e rim e n t. The c o r re c t e stim a tio n o f th e v e rtic a l and h o rizo n ta l m erid ian s, d e lin e a tin g th e e a rly visu al areas, a llo w e d ass ign m en t o f resp on ses and ROls w ith in c y to a rc h ite c to n ic areas p u re ly o n th e basis o f the
to m o g r a p h ic source analysis o f IVIFG signals at least fo r V I , V2 and V3. T he resu ltin g spatial d istrib u tio n and te m p o ra l s eq u en ce o f e v o k e d resp on ses are in lin e w it h a fe e d fo r w a r d s w e e p o f a c tiv ity th rou gh th e visu al h iera rc h y (L a m m e and R oelfsem a, 2 0 0 0 ).
Earlier studies
P a tie n t G Y has b e e n e x a m in e d in n u m erou s p re v io u s studies. T h e resu lts w e re p o rte d h ere are la r g e ly c o n s is te n t w it h th e se p re v iou s find in gs, a lth ou gh a d e ta ile d c o m p a ris o n has n o t b e e n a l w a y s p ossib le, e ith e r becau se d iffe r e n t s tim u li and p ro to c o ls h ave b e e n used o r b ecau se p re v io u s te c h n iq u e s d id n o t p r o v id e results w it h c o m p a ra b le tim in g and spatial accuracy. In the present study w e have opted for stimuli know n to produce a strong response in V I so as to m in im ize the chances o f m issing w e a k activations initiated in “ islands o f preserved c ortex" in early retinotopic areas that have been suggested as possible explanation o f blindsight in several patients. Our results found no such evidence, thus strengthening the conclusions from earlier studies o f G Y (K en trid ge et al., 1997; W eiskrantz, 2009a). The identification o f activity in the dam aged hem isphere in high-level areas is thus even m ore convincing because the stimuli w e used do not usually g iv e strong responses in these areas. The activation id enti fied close to the IVIT -I-/V5 in our study is consistent w ith other studies using PFT (Barbur et al., 1993), IVIFG (H o llid a y e t al., 1997) and fMRI (G oeb el e t al., 2001) in the same patient. In the early PFT (Barbur et al., 1993) and MFG (H ollid a y e t al., 1997) studies the actual tasks and stimuli w e re rather d ifferen t and no coordinates are available for d e tailed com parison. Based on the Talairach coordinates (Talairach and Tournoux, 1988) the MT-l-A/5 activity found in the fMRI study (G oeb el e t al., 200 1) is ~ 1.4 cm from our M T -I-A/5 ROl, and the activity
A .
144 156 162 ms
102 156 162 ms
Fig. 7. Activity during the second visual processing phase, at three latencies. For each tim e instant the activations (S N R > 0.2) in response to “ unseen” (A B and C) and “seen” (D, E and F) stimuli are shown on the lateral, m edial and ventral aspects o f inflated cortical surfaces o f the ipsilesional left (A, B and C) and contralesional right (D, E and F) hemispheres respectively. Putative V4 and V5 areas obtained from the 50% cytoarchitectonic probabilistic maps are also shown on the lateral (V 5 ) and ventral (V 4 ) view s o f each figure. The black patches sh ow the lesioned portion o f the cortex. The insets indicated by black arrows highlight the responses in the early visual cortex. They show the zoom ed and slightly rotated view s o f the inflated cortical surface areas encompassed in corresponding black rectangles. A Stimuli presented in the UR quadrant. B. On the RHM. C. In the LR quadrant. D. In the UL quadrant. E. On the LHM. F. In the LL quadrant
G oebel e t al. identified in response to coloured im ages o f natural objects is -1 .3 cm from our M OG ROl.
It is w o rth com paring our results w ith those o f a com bined fM R l and MEG study in another hem ianopic patient w ith cortical blindness, ad m ittedly o f much you n ger age (Schoenfeld e t al., 2002). The activity center for V5 from this study w as -2 .2 cm a w a y from ours and the on e for LOT -0 .9 cm from our MOG activity. The tim ing o f the peak
activity elicited b y stimuli in the blind hem ifield was not quoted, but the beginning o f the activity was similar to the on e observed in our study, as w as the observation that “ ... e vok ed activity occurred earlier in the h igher-tier visual areas V4/V8 and V5 than in the low e r-tie r areas V2/V3 adjacent to the lesion."
Although the experim ents discussed above differ in im portant ways to the on e w e did here, the overall consistency o f the results across studies is
UR
A
LR
100 200 300
Time (ms)
Fig. 8. SP tim e courses for estim ated V2 ROIs in GY. Each figure displays tim e course for
one quadrant o f visual field (printed on its upper left part) fo r the corresponding ROl.
important, because it suggests tiiat an access to iiigiier level visual cortex, is available from the blind field and is not dependent on V I. Recent e vi dence from animal and human neuroimaging studies suggests the exis tence o f at least tw o separate extra-geniculostriate pathways bypassing V I and linked to functional and behavioral evidence o f blindsight. One pathw ay involves direct connections b etw een the intercalated layers o f the lateral geniculate nucleus o f the thalamus and area M T -l-A^S, which is especially involved in m ovem en t perception (Bridge e t al., 2008; Schmid e t al., 2010). The other pathw ay involves a disynaptic pathway
100 200
Time (ms)
Fig. 9. SP tim e courses for the three post-retinotopic ipsilesional ROIs in GY. Tim e
courses fo r the three ROIs are superimposed (red - left MOG BA19, blue - left MT-H/V5, green - left MTG BA37). Each figure displays tim e courses for one location o f the visual field, which is specified in its upper left part (UVM, U R RHM, LR, LVM, LL, LHM and UL) and corresponds to its schematic location w ithin the figure.
connecting the superior colliculus to extrastriate occipito-tem poral areas via the visual pulvinar, particularly active in visual-m otor integra tion and “attention blindsight" (Lyon e t al., 2010; Tam ietto et al., 2010). The spatiotemporal precision o f our results supports and extends these previous findings show ing that, if this is indeed the w a y the stimulus e f fect reaches the blind hemisphere, this extra-geniculostriate access is ac tivated during the second phase o f visual processing and it spreads back into the earlier visual areas in a non-retinotopic manner and w ithou t pre cise tim e sequencing. W e did not attem pt to identify the subcortical origin o f this extrastriate activity because the MEG hardware w e used m ay not be v e ry sensitive to deep sources. Another possible access to higher level visual cortex m ay be from neurons w ith v e ry large receptive fields dispersed in early retinotopic cortex (e.g. in prostriata or V2 areas). These neurons m ay not form topographically w ell-d efin ed groupings o f neurons and hence produce no measurable MEG signal, but they m ay converge onto w ell-d efin ed circumscribed regions in the high-level visual areas and thus produce the activations w e have identified. The appear ance o f these activations in the second phase o f processing m ay be due to a com bination o f the contribution from earlier visual areas and directly from subcortical routes, as both inputs m ight be needed to give rise to a measurable signal.
Data analysis m ethodology
In the present study w e used tom ographic source analysis (M ET) (loannides et al., 1990; M oradi et al., 2003; Papadelis et al., 2009; Poghosyan and loannides, 2007; Taylor et al., 1999) o f single trial MEG signals to identify the precise spatiotemporal sequence o f visual cortical responses follow in g presentation o f simple checkerboard patterns. MET was used routinely to map activity in the brain o f individual subjects for alm ost e very MEG multichannel hardware, from the early systems cover ing on ly part o f the head (Ribary e t al., 1991) to m ost o f the w h ole head systems used today, including the tw o used in the current study. Although MET can reliably recover reasonably deep sources (loannides et al., 2005), in the current study w e focused on ly on relatively superficial neural sources w h ere the sensitivity o f both MEG hardware is excellent. Separate experim ents have demonstrated that activity w ithin V I can be reliably extracted and differentiated from activity in extrastriate areas w h en the V I- V 2 border is available from an independent fMRl measurement (M oradi et al., 2003). M ore recently w e also show ed that inform ation from cytoarchitectonic probability maps can be com bined w ith the to m o graphic estimates (Papadelis e t al., 2011), and that the spatiotemporal evolution o f activity elicited b y simple stimuli can be reliably and repro- ducibly extracted in different cytoarchitectonic areas (Poghosyan and loannides, 2007). Here, the analysis o f simulated data has demonstrated that the activity in locations w ithin V I and V2 can be disentangled using the tw o different MEG systems. In this w ork w e have also extended the m ethodology so as to define the borders o f visual areas and map the activity w ithin each visual area from the MEG data alone. W e tested the n e w capability using fMRl V I- V 2 borders for normal subjects and the cytoarchitectonic probabilistic maps for all subjects. W h ile such a delinea tion o f early visual cortical areas using MEG is a v e ry valuable capability, m ore detailed tests and validation w ith fMRl-based retinotopic maps are necessary to confirm accuracy and reliability o f the procedure. Taken together our results demonstrate the effectiveness o f the m ethod o lo g y for both MEG systems used in this study and the results o f the anal ysis o f the real MEG data underscore their overall consistency w ith know n properties o f the visual system.
D ifferen t M EG systems
Could the use o f d iffe re n t MEG hardw ares in G Y and norm al sub jects introduce spurious effects in our results? T he rationale o f our analysis is sim ple: our m ain com parison is b e tw e e n the intact and d am aged hem ispheres o f GY. Eor this critical com parison, the issue o f d ifferen t MEG facilities d oes n ot apply. H o w e ve r, w e first tested
Left MTG BA37
Left MT+A/5
Left MOG BA19
-100 0 100 200 300 -100 0 100 200 300
0 100 200
Time (ms)
Fig. 10. SP tim e courses in a typical control subject fo r three ROls corresponding to GY's
three post-retinotopic ipsilesional ROls shown in Fig. 8. Tim e courses for the three ROls are superimposed (red — left MOG BAl 9, blue — left MT-l-A^5, green — left MTG BA37). Each figure displays tim e courses for one location o f the visual field, which is specified in its upper left part (UVM, UR, RHM, LR LVM, LL, LHM and UL) and corresponds to its schematic location w ithin the figure. The results for stimuli presented at 9° eccentricity are shown.
the hypothesis that the neural processing in the intact hem isp h ere o f G Y is consistent w ith w h a t w e have ob served in our d etailed study in n orm al subjects (a lb e it w ith an oth er IVIEG h ard w a re). Indeed, w e found v e r y sim ilar spatiotem p oral response patterns in the intact h em isp h ere o f G Y and in health y subjects. If the result w as d ifferen t th en one could ju stifia b ly argue that these d ifferen ces w e r e due to the d ifferen ces in the hardw are. The results w e r e h o w e v e r sim ilar and this, if anything, argues that th ere w as no u n expected e ffe c t in trod u ced b y the use o f d iffe re n t IVIEG system s on our results. The sim ilarity o f the ac tiv ity patterns b e tw e e n the intact hem isp h ere o f G Y and h ealth y subjects, and th e consistency w ith the previous literature show s that any possible spurious e ffe c t o f d ifferen t MEG hardw ares is n egligible. The key com parison b e tw e e n the intact and d am aged h em isp h ere o f G Y is, o f course, d on e w ith the same system.
A t the tim e o f the experim ent w ith GY the MEG laboratory at RIKEN Brain Science Institute, w h ere the experim ents w ith normal subjects have been conducted, had been decommissioned. Therefore the experi m ent w ith GY was conducted in a different laboratory at ULB-Hopital Erasme using identical parameters for the stimuli, except for small d iffer ences to ensure that stimuli presented in the right hem ifield w e re entirely w ithin the blind field o f GY. All data w e re analyzed using a robust m eth od ology (Data analysis m eth odology section) that was successfully ap plied before to data from different MEG systems, including studies w ith the tw o MEG systems used in the current study and studies on the same subjects w ith different MEG hardware (loannides et al., 2002). Part o f the preparatory w ork for the current study included com puter simulations (Com puter simulations and Analysis o f simulated data sec tions, and Supplementary m aterial) and analysis o f magnetic phantom measurements from the ULB-Hopital Erasme MEG hardware (data not presented here).
O ther brain activations
In addition to the identified ROls that w e re robustly activated b y our stimuli, tw o additional labile interm ittent sources w e re evid en t in the in tact hem isphere o f GY: one source along the parieto-occipital sulcus (see Fig. 3C at 60 m s) and another in ‘d eep ’ subcortical structures (see Fig. 3C at 66 ms). An early, concurrent w ith V I, interm ittent activity along the parieto-occipital sulcus has been reported also earlier (P ortin e t al., 1998; Tzelepi et al., 2001; Vanni e t al., 2001). Studies have suggested this area to be the human hom ologue o f m onkey's V6 and reflect an automatic activation o f a part o f the attention- or visuom otor-related networks (Vanni et al., 2001).
The d eep source on the o th er hand m ay be an artifact o f the source localization process, or a reflection o f a true subcortical source. The form er exp lan ation w e consider unlikely, because the m e th o d o lo g y w e used here d oes n ot usually have such artifacts. It solves the inverse p rob lem using five separate source spaces (s e e T om ograp h ic source analysis section ). T he solutions are then com b in ed to g e th e r into a final source space that covers th e entire brain. T he use o f regu lariza tion in solvin g the inverse p ro b le m reduces such ‘g h ost’ images, but e ve n in the case w h e re the regu larization is n ot enou gh to elim in ate such sources, th e y w ill ap p ear at d ifferen t locations in each source space, and w ill be h ea vily p en alized w h e n th e solutions fro m d iffe r en t source spaces are com bined. N evertheless, w e cannot fu lly e x clude such possibility.
The marked asym m etry b etw e en the brain responses to upper and lo w e r visual field stimuli observed in the current study has been reported earlier in norm al subjects (Fylan e t al., 1997; Portin e t al., 1999; T zelep i e t al., 2001) as w e ll as in G Y (d e G elder et al., 2001). Evi dence suggests that upper and lo w e r visual fields have d ifferen t func tional roles, w ith lo w e r field superiority found in m a jority o f tasks (Bilodeau and Faubert, 1997; Danckert and Goodale, 2001; Genzano et al., 2001; Rubin e t al., 1996; Skrandies, 1987). H ow ever, upper field ad vantages in a num ber o f tasks have also been reported (G oldstein and B abkoff 2001; Zhou and King, 2002). M oreover, receptors and ganglion cells are denser in the human upper than lo w e r h em iretina (Curcio and Allen, 1990; Curcio e t al., 1990), ensuing an anatom ical o verrep resenta tion o f the lo w e r fields in the retina. This visual field specialization has been attributed to the adaptation o f the human visual system to the e n v iron m en t in w h ich w e live (Previc, 1990; Skrandies, 1987).
Conclusions
Our results sh o w that the response to “u nseen" stim uli in a corti cally blind p atient w ith V I lesion is d ela yed in tim e, d oes not adhere to classic retin otop ic orga nization and appears first on the corte x in h igh -lev e l extrastriate dorsal areas, w hereas the spatiotem p oral re sponses to the intact hem isp h ere are sim ilar to the standard pattern ob served in h ealth y subjects.
Disclosure statement
Authors d eclare no c on flict o f interest.
Role o f the funding source
The funding agencies did n ot p lay any role in the study design, in the collection, analysis, and in terp retation o f data, in the w ritin g o f the re p o rt o r in the decision to subm it the pap er for publication.
S upplem entary data to this article can be found on line at http:// dx.doi.org/10.1016/j.neuroim age.2012.07.058.
Aclmowledgment
W e thank patient GY for participating in the study. The experim ents w ith the three control subjects w e re perform ed at the Laboratory for
