Title: Decoding Seen and Attended Edge Orientation and Motion Direction from the Human Brain Activity Measured by functional Magnetic Resonance Imaging (fMRI)
1Decoding Seen and Attended Edge Orientation and
Motion Direction from the Human Brain Activity
Measured by functional Magnetic Resonance Imaging
(fMRI)
- Presented by Arash Ashari
- Kamitani, Y., and Tong, F. (2005). Decoding the
visual and subjective contents of the human
brain. Nat. Neurosci. 8, 679685. - Kamitani, Y., Tong, F. (2006). Decoding seen and
attended motion directions from activity in the
human visual cortex. Current Biology, 16
1096-1102.
2Outline
- functional Magnetic Resonance Imaging (fMRI)
- Visual Cortex
- Orientation Decoder
- Direction Decoder
- Data Analysis
- fMRI- Accomplishments and Future works
3Outline
- functional Magnetic Resonance Imaging (fMRI)
- Visual Cortex
- Orientation Decoder
- Direction Decoder
- Data Analysis
- fMRI- Accomplishments and Future works
4MRI Magnetic Resonance Imaging
- Using Nuclear Magnetic Resonance (NMR)
technology, magnetic field influence the nucleus
of hydrogen. - The MRI transmit Radio Frequency (RF) wave to the
nucleus, and measures changes in magnetic field.
- The RF changes according to the chemical
structure of the tissue. - The computerized images give a detailed
anatomical view of the organ.
- High Spatial Resolution
- Used in brain structure research and
localization of brain-tumor
http//www.bm.technion.ac.il/courses/335014/projec
ts04/mid_term_presentations_05/Mapping_visual_cort
ex.ppt
5fMRI Functional MRI
- The new neuroimaging method for probing the
intact human brain. - The technique is based on
- In neural activity an additional supply of
oxygenated blood is delivered. - Oxygenated Hemoglobin (Hb) is magnetically
transparent (diamagnetic). - Deoxygenated Hb is not transparent
(paramagnetic). - The change in the ratio of oxygenated to
deoxygenated can be detected in the MR signal. - This signaling mechanism is the Blood Oxygen
Level-Dependent (BOLD)
6fMRI-Characteristics
- It measures changes in the subject in real time,
without external Indicator. - The exam can be repeated many times, without any
harm. - High Spatial Resolution mm
- Relatively High Temporal Resolution Sec
- High SNR allows to measure the size of
differences, not just their presence or absence.
7Outline
- functional Magnetic Resonance Imaging (fMRI)
- Visual Cortex
- Orientation Decoder
- Direction Decoder
- Data Analysis
- fMRI- Accomplishments and Future works
8Visual Cortex
- The term visual cortex refers to the primary
visual cortex/V1 and extra-striate visual
cortical areas such as V2, V3, V4, and V5/MT. - The primary visual cortex, V1, receives
information directly from the lateral geniculate
nucleus. V1 transmits information to two primary
pathways, called the dorsal stream and the
ventral stream.
9Primary Visual Pathways
- The dorsal stream begins with V1, goes through
Visual area V2, then to the dorsomedial area and
Visual area MT and to the posterior parietal
cortex. The dorsal stream, sometimes called the
"Where Pathway", is associated with motion,
representation of object locations, and control
of the eyes and arms. - The ventral stream begins with V1, goes through
visual area V2, then through visual area V4, and
to the inferior temporal cortex. The ventral
stream, sometimes called the "What Pathway", is
associated with form recognition and object
representation. It is also associated with
storage of long-term memory.
http//en.wikipedia.org/wiki/Visual_cortex
10Visual Perception
- It is commonly assumed that human visual
perception is based on the neural coding of
fundamental features, such as Orientation, Color,
Motion and so forth. - So it can be hypothesized that functional
neuroimaging (fMRI) identifies brain areas that
show robust responses to visual orientation and
motion.
11Outline
- functional Magnetic Resonance Imaging (fMRI)
- Visual Cortex
- Orientation Decoder
- Direction Decoder
- Data Analysis
- fMRI- Accomplishments and Future works
12Orientation Decoder
- Subjects views one of eight possible stimulus
orientations while activity is monitored in early
visual areas (V1-V4 and MT) using standard fMRI
procedures (3T MRI scanner, spatial resolution
333 mm). For each 16-s 'trial' or stimulus
block, a square-wave annular grating is presented
at the specified orientation (0, 22.5, ...,
157.5), and flashes on and off every 250 ms with
a randomized spatial phase to ensure that there
is no mutual information between orientation and
local pixel intensity.
13Orientation decoding accuracy
- fMRI activity patterns in the human visual cortex
are sufficiently reliable to predict what
stimulus orientation the subject is viewing on
individual trials. - Ensemble fMRI activity in areas V1/V2 led to
precise decoding of which of the eight
orientations the subject saw on individual
stimulus trials. - Root mean squared error (RMSE) between the true
and the predicted orientations - 17.9, 21.0, 22.2 and 31.2, respectively for
subjects S1-S4
14Orientation decoding accuracy across visual areas
- The ability to extract robust orientation
information from ensemble fMRI activity allows us
to compare orientation selectivity across
different human visual areas. Orientation
selectivity is most pronounced in early areas V1
and V2, and declines in progressively higher
visual areas. Unlike areas V1 through V4, human
area MT showed no evidence of orientation
selectivity consistent with the idea that this
region is more sensitive to motion than to
stimulus form.
15Source of orientation Information
- The orientation preference of individual voxels
on the flattened surface of left ventral V1 and
V2 for subjects S2 and S3. - Voxel colors depict the orientation detector for
which each voxel provides the largest weight.
16Mind-Reading of Attended Orientation
- The robust effects found in V1 and V2 suggest
that top-down voluntary attention acts very early
in the processing stream to bias
orientation-selective signals when two competing
stimuli are entirely overlapping.
17Outline
- functional Magnetic Resonance Imaging (fMRI)
- Visual Cortex
- Orientation Decoder
- Direction Decoder
- Data Analysis
- fMRI- Accomplishments and Future works
18Direction decoder
- The decoder receives fMRI voxel intensities,
averaged for each 16-s stimulus block, as inputs. - The next layer consisting of linear ensemble
direction detectors calculates the weighed sum
of voxel inputs. - Voxel weights are optimized using a statistical
learning algorithm applied to independent
training data, so that each detectors output
become larger for its direction than for the
others. - The direction of the most active detector is used
as the prediction of the decoder.
19Comparison of Orientation and Direction
Selectivity
20Mind-Reading of Attended Direction
- Feature-based attention can alter the strength
of direction-selective responses throughout the
visual pathway, with top-down bias effects
emerging at very early stages of visual
processing. ( Attention should bias the pattern
of neural activity to more closely resemble the
activity pattern that would be induced by the
attended feature alone.)
21Outline
- functional Magnetic Resonance Imaging (fMRI)
- Visual Cortex
- Orientation Decoder
- Direction Decoder
- Data Analysis
- fMRI- Accomplishments and Future works
22Data Analysis
- Three dimensional (3D) motion correction
- Linear trend removal (using automated image
registration software) - Sorting the voxels according to the responses to
the visual areas. - Shifting the data 4 s to account for the
hemodynamic delay - Averaging the fMRI signal intensity of each voxel
for each 16 s stimulus trail. - Normalization relative to average of the entire
time course within each run - Labeling the activity patterns according to
corresponding stimulus Orientation or Direction. - Classification (using cross-validation for train
and test)
23Classification
- The classification is done by a linear ensemble
Orientation/Direction detector - ?k preferred Orientation/direction
- x(x1, x2,..., xd) voxel inputs
- Linear detector function
- where wi is the weight of voxel i, and w0 is
the bias
24Classification (con)
- Linear Support Vector Machine (SVM) is used to
calculate a linear discriminant function for the
pairs of all orientation/direction -
-
- Then the pairwise discriminant functions
comparing ?k and the other directions are simply
added to yield the linear detector function
25Outline
- functional Magnetic Resonance Imaging (fMRI)
- Visual Cortex
- Orientation Decoder
- Direction Decoder
- Data Analysis
- fMRI- Accomplishments and Future works
26fMRI- Accomplishments and Future works
- Identification of the position of several
retinotopically organized visual areas. - Measurement of the retinotopic organization
within these areas. - Identification of the location of
orientation/motion-sensitive region. - Measurement of the responses associated with
contrast, color and motion. - Measurement of localized deficits in activity in
subjects with cortical damage. - Measurement of the effects of attentional
modulation on visually evoked responses.
High inter-subject variability of functional
activity and sheer dimensionality of fMRI data
are the two major problems in fMRI data
classification.
27Another Active Project on Mind-Reading
- Classifying cognitive processes based on fMRI
images How can we use statistical machine
learning methods to classify the hidden cognitive
process of a human subject, based on their
observed fMRI data? - In this fMRI study Mitchll et al trained their
algorithms to decode whether the words being read
by a human subject are about tools, buildings,
food, or several other semantic categories. The
trained classifier is 90 accurate, for example,
discriminating whether the subject is reading
words about tools or buildings.
http//www.cs.cmu.edu/afs/cs/project/theo-73/www/i
ndex.html
28Thank you
Much remains to be learned about how the human
brain represents the basic attributes of visual
experiences.