WEBVTT
Kind: captions
Language: en

00:00:13.660 --> 00:00:18.260
This is the algorithm the flow chart.

00:00:19.160 --> 00:00:22.460
Including the ADC map

00:00:22.460 --> 00:00:27.000
and the DWI image registration.

00:00:27.320 --> 00:00:33.700
The ADC map was registered to the corresponding DWI

00:00:33.700 --> 00:00:40.340
by a rigid registration (translation and rotation) and a trilinear interpolation

00:00:40.340 --> 00:00:44.760
based on the normalized mutual information method

00:00:44.760 --> 00:00:49.580
to correct for differences due to head movements.

00:00:49.580 --> 00:00:54.900
The DWI and registered ADC map

00:00:54.900 --> 00:01:05.700
were normalized so that their intensities were both distributed in a standardized range.

00:01:05.700 --> 00:01:08.140
From zero to one.

00:01:08.140 --> 00:01:13.200
The program we used to run this step

00:01:13.200 --> 00:01:20.180
was the Statistical Parametric Mapping it's meaning SPM.

00:01:20.180 --> 00:01:28.020
And the next step is extracting from the brain mask from the whole-brain DWI.

00:01:28.020 --> 00:01:36.760
The brain mask was extracted from the whole-brain DWI based on the estimation

00:01:36.760 --> 00:01:45.580
of the inner and outer skull surfaces
by using BET (Brain Extraction Tool),

00:01:45.580 --> 00:01:53.040
Step 3 is about preclustering elimination.

00:01:53.040 --> 00:01:59.580
The histogram of DWI within the brain mask

00:01:59.580 --> 00:02:04.980
was smoothed by a third-order
moving-average filter.

00:02:04.980 --> 00:02:10.120
The peak of the smoothed histogram was identified.

00:02:10.120 --> 00:02:15.900
The normalized intensity, denoted by I peak,

00:02:15.980 --> 00:02:22.960
corresponding to this histographic peak was used as a threshold value.

00:02:23.280 --> 00:02:30.120
The voxels with normalized intensities lower than
or equal to I peak

00:02:30.120 --> 00:02:35.660
would be eliminated from further processing.

00:02:35.660 --> 00:02:40.020
Step 4 is about FCM clustering.

00:02:40.020 --> 00:02:43.880
The remaining voxels after the previous step

00:02:43.880 --> 00:02:52.840
were divided into 50 clusters by an unsupervised classification

00:02:52.840 --> 00:02:58.900
with the conventional FCM clustering algorithm.

00:02:58.900 --> 00:03:06.140
Step 5 is about skimming the clusters from candidate voxels.

00:03:06.140 --> 00:03:14.280
The clusters with mean normalized intensity larger than the normalized intensity

00:03:14.280 --> 00:03:20.060
of the histographic peak plus 0.2,

00:03:20.060 --> 00:03:25.620
that is, I peak plus 0.2, were selected.

00:03:25.620 --> 00:03:32.500
The voxels belonging to these selected clusters would be treated as candidate voxels

00:03:32.500 --> 00:03:35.580
of infarct in the next step.

00:03:35.580 --> 00:03:44.380
Next step is eliminating labels with insufficient intensity.

00:03:44.380 --> 00:03:54.120
Each cluster of candidate voxels of infarct was further divided into one or several labels,

00:03:54.120 --> 00:04:00.900
at most the same number as the voxel number in that cluster.

00:04:00.900 --> 00:04:04.600
A label comprised connected voxels.

00:04:04.600 --> 00:04:13.760
The labels with average normalized intensity lower than or equal to the threshold value (I peak + 0.2)

00:04:13.760 --> 00:04:18.120
were eliminated from further processing.

00:04:18.120 --> 00:04:23.540
The next step is eliminating candidate labels

00:04:23.540 --> 00:04:26.980
due to magnetic inhomogeneity.