Contains classical quality metrics for assessing the quality of a segmentation per label relative to a ground truth.
Information about the measurement named measurement can be accessed by using the method
measurementInformation() which returns a
FieldInformation
Information about the measurement named measurement can be accessed by using the property
measurementInformation which returns a
FieldInformation
Information about the measurement named measurement can be accessed by using the property
measurement_information which returns a
FieldInformation
Object members
Measurement name
Description
Element type
Indexing
Physical Information
labelValue
The label value corresponding to the measurement current index.
Unsigned integer
[time, label]
INTENSITY
truePositive
The number of positive results in input image that are defined as such in the ground truth image (TP).
Unsigned integer
[time, label]
COUNT
falsePositive
The number of positive results in input image that are not defined as such in the ground truth image (FP).
Unsigned integer
[time, label]
COUNT
trueNegative
The number of negative results in input image that are defined as such in the ground truth image (TN).
Unsigned integer
[time, label]
COUNT
falseNegative
The number of negative results in input image that are not defined as such in the ground truth image (FN).
Unsigned integer
[time, label]
COUNT
sensitivity
The proportion of actual positives that are correctly identified as such in the ground truth image.
This metric is also called recall or true positive rate. $$\frac{TP}{TP+FN}$$
Floating point
[time, label]
COEFFICIENT
specificity
The proportion of actual negatives that are correctly identified as such in the ground truth image.
$$ \frac{TN}{TN+FP}$$
Floating point
[time, label]
COEFFICIENT
precision
The fraction of relevant results (true positives) among positive results (true and false positives).
$$\frac{TP}{TP+FP}$$
Floating point
[time, label]
COEFFICIENT
accuracy
The percentage of correct classification among all results.
$$ \frac{TP+TN}{TP+FN+TN+FP} $$
Floating point
[time, label]
PERCENTAGE
dice
The Sorensen Dice coefficient, comparing the similarity of two samples. This metric is also called F1 score.
$$ \frac{2TP}{2TP+FP+FN} $$
Floating point
[time, label]
COEFFICIENT
jaccard
The mean Intersection-Over-Union metric.
$$ \frac{TP}{TP+FP+FN} $$
Floating point
[time, label]
COEFFICIENT
Object methods
Method
Description
void toDataFrame()
Convert the measurement to an IOLink.DataFrame
One or more "index" columns will be added at the beginning of the dataframe to identify
the different elements. For instance:
index [label]: The index of the label. This index systematically starts from 0
and is the label value minus 1 when all label values are represented.
When there are some missing label values, the corresponding label value cannot be
directly deduced from this index.
index [time]: The sequence index of a time series or of an image stack.
Method
Description
void ToDataFrame()
Convert the measurement to an IOLink.DataFrame
One or more "index" columns will be added at the beginning of the dataframe to identify
the different elements. For instance:
index [label]: The index of the label. This index systematically starts from 0
and is the label value minus 1 when all label values are represented.
When there are some missing label values, the corresponding label value cannot be
directly deduced from this index.
index [time]: The sequence index of a time series or of an image stack.
Method
Description
void to_data_frame()
Convert the measurement to an IOLink.DataFrame
One or more "index" columns will be added at the beginning of the dataframe to identify
the different elements. For instance:
index [label]: The index of the label. This index systematically starts from 0
and is the label value minus 1 when all label values are represented.
When there are some missing label values, the corresponding label value cannot be
directly deduced from this index.
index [time]: The sequence index of a time series or of an image stack.