Upload registry, index and card
Browse files- README.md +3 -17
- index.parquet +2 -2
README.md
CHANGED
|
@@ -35,14 +35,13 @@ can select what you need before downloading anything.
|
|
| 35 |
evaluating a gaze probe. Same format; `labels` is simply present. These are
|
| 36 |
the `dsL##_*` folders, so `dsL*/*.h5` selects them without opening a file.
|
| 37 |
|
| 38 |
-
`dsL01`–`dsL06` come from the DeepMReye 1.0 training sets. `dsL07`
|
| 39 |
ingested from OpenNeuro datasets that recorded eye tracking during the scan:
|
| 40 |
|
| 41 |
| folder | source | n | paradigm |
|
| 42 |
|---|---|---|---|
|
| 43 |
| `dsL07_deepmreye_calib` | ds006833 | 15 | fixation / pursuit / free viewing |
|
| 44 |
-
| `
|
| 45 |
-
| `dsL09_fearlearning` | ds001242 | 52 | fear learning |
|
| 46 |
|
| 47 |
### Gaze/BOLD alignment
|
| 48 |
|
|
@@ -53,16 +52,6 @@ to absorb an error. Each file records how its time origin was recovered
|
|
| 53 |
(`gaze_anchor`: a BIDS `StartTime`, a scanner-trigger column, or a sync message
|
| 54 |
in the tracker stream) and any residual offset applied (`gaze_time_offset`).
|
| 55 |
|
| 56 |
-
Two datasets needed a sub-TR correction, estimated per dataset (one scalar) and
|
| 57 |
-
consistent across participants: `dsL08` −0.75 s (15/15 subjects agreed on the
|
| 58 |
-
sign; worth +0.12 mean r) and `dsL09` +0.50 s (9/10; +0.06). Those two scalars
|
| 59 |
-
were estimated *with* a decoder, so absolute decodability numbers for them
|
| 60 |
-
should say so; comparisons between methods are unaffected.
|
| 61 |
-
|
| 62 |
-
One further dataset (ds007532) was extracted and then **rejected**: its
|
| 63 |
-
per-run timing could not be recovered from anything it ships, and no
|
| 64 |
-
dataset-level offset exists. It is not in this repo.
|
| 65 |
-
|
| 66 |
## Processing
|
| 67 |
|
| 68 |
Each functional run was coregistered to the DeepMReye template with ANTs
|
|
@@ -83,12 +72,9 @@ and the gaze go out of alignment.
|
|
| 83 |
|---|---|
|
| 84 |
| `dsL01`–`dsL06` | degrees of visual angle |
|
| 85 |
| `dsL07_deepmreye_calib` | degrees of visual angle |
|
| 86 |
-
| `dsL08_studyforrest_movie` | degrees of visual angle |
|
| 87 |
| `dsL11_backtothefuture` | degrees of visual angle |
|
| 88 |
-
| `dsL09_fearlearning` | screen pixels |
|
| 89 |
-
| `dsL12_rest` | screen pixels |
|
| 90 |
|
| 91 |
-
The source papers were checked for each dataset. `dsL07`
|
| 92 |
|
| 93 |
Pearson correlation is invariant to the difference. **Training is not**: if you
|
| 94 |
fit one readout over several datasets pooled, standardise the target per dataset
|
|
|
|
| 35 |
evaluating a gaze probe. Same format; `labels` is simply present. These are
|
| 36 |
the `dsL##_*` folders, so `dsL*/*.h5` selects them without opening a file.
|
| 37 |
|
| 38 |
+
`dsL01`–`dsL06` come from the DeepMReye 1.0 training sets. `dsL07` and `dsL11` were
|
| 39 |
ingested from OpenNeuro datasets that recorded eye tracking during the scan:
|
| 40 |
|
| 41 |
| folder | source | n | paradigm |
|
| 42 |
|---|---|---|---|
|
| 43 |
| `dsL07_deepmreye_calib` | ds006833 | 15 | fixation / pursuit / free viewing |
|
| 44 |
+
| `dsL11_backtothefuture` | ds006642 | 4 | movie (backtothefuture) |
|
|
|
|
| 45 |
|
| 46 |
### Gaze/BOLD alignment
|
| 47 |
|
|
|
|
| 52 |
(`gaze_anchor`: a BIDS `StartTime`, a scanner-trigger column, or a sync message
|
| 53 |
in the tracker stream) and any residual offset applied (`gaze_time_offset`).
|
| 54 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 55 |
## Processing
|
| 56 |
|
| 57 |
Each functional run was coregistered to the DeepMReye template with ANTs
|
|
|
|
| 72 |
|---|---|
|
| 73 |
| `dsL01`–`dsL06` | degrees of visual angle |
|
| 74 |
| `dsL07_deepmreye_calib` | degrees of visual angle |
|
|
|
|
| 75 |
| `dsL11_backtothefuture` | degrees of visual angle |
|
|
|
|
|
|
|
| 76 |
|
| 77 |
+
The source papers were checked for each dataset. `dsL07` and `dsL11` document geometry (display size and viewing distance) to convert to degrees of visual angle.
|
| 78 |
|
| 79 |
Pearson correlation is invariant to the difference. **Training is not**: if you
|
| 80 |
fit one readout over several datasets pooled, standardise the target per dataset
|
index.parquet
CHANGED
|
@@ -1,3 +1,3 @@
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:
|
| 3 |
-
size
|
|
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:ff5272d57fc6f28c4d0a36300f10ae554740dc23e3c85a880fabc5fe4781d78c
|
| 3 |
+
size 84746
|