US 10,853,695 B2Granted

Method and system for cell annotation with adaptive incremental learning

AssigneeKonica Minolta Laboratory U.S.A., Inc.

InventorsYongmian Zhang, Jingwen Zhu

Life of this patent

Estimating the term…

Loading patent…

Abstract

A method, a computer readable medium, and a system for cell annotation are disclosed. The method includes receiving at least one new cell image for cell detection; extracting cell features from the at least one new cell image; comparing the extracted cell features to a matrix of cell features of each class to predict a closest class, wherein the matrix of cell features has been generated from at least initial training data comprising at least one cell image; detecting cell pixels from the extracted cell features of the at least one new cell image using the predicted closest class to generate a likelihood map; extracting individual cells from the at least one cell image by segmenting the individual cells from the likelihood map; and performing a machine annotation on the extracted individual cells from the at least one new cell image to identify cells, non-cell pixels, and/or cell boundaries.

Claims17

Dependent claims

A method for cell annotation, the method comprising:

receiving at least one new cell image for cell detection;
extracting cell features from the at least one new cell image;
comparing the extracted cell features to a matrix of cell features of each class to predict a closest class, wherein the matrix of cell features has been generated from at least initial training data comprising at least one cell image, wherein the comparing the extracted cell features from the at least one new cell image to the matrix of cell features is performed by a similarity matching, the similarity matching comprising:
performing a suitable distance measurement between histograms using:
D
⁡
(
H
⁡
(
a
)
,
H
⁡
(
b
)
)
=
∑
1
k
⁢
(
1
-
1
+

min
⁡
(
h
i
⁡
(
a
)
,
h
i
⁡
(
b
)
)

1
+

max
⁡
(
h
i
⁡
(
a
)
,
h
i
⁡
(
b
)
)

+

min
⁡
(
h
i
⁡
(
a
)
,
h
i
⁡
(
b
)
)

)
where H(a) is a histogram for image a and H(b) is a histogram for image b with k histogram bins;
mapping colors in the image into a discrete color space containing k colors, then the color histogram H(a) is a vector [h1(a), . . . , hk(a)], and H(b) is a vector [h1(b), . . . , hk(b)];
and
wherein if D(H(a), H(b)) is greater than a predefined threshold, the image belongs to an unknown class, and if the less than the predefined threshold, then the image belongs to a known class;
detecting cell pixels from the extracted cell features of the at least one new cell image using the predicted closest class to generate a likelihood map;
extracting individual cells from the at least one cell image by segmenting the individual cells from the likelihood map;
performing a machine annotation on the extracted individual cells from the at least one new cell image to identify cells, non-cell pixels, and/or cell boundaries;
calculating a confidence level for the machine annotation on the extracted individual cells from the at least one new cell image;
and
modifying the machine annotation if the confidence level is below a predetermined threshold.

97 more elements

A method for cell annotation, the method comprising:

receiving at least one new cell image for cell detection;
extracting cell features from the at least one new cell image;
comparing the extracted cell features to a matrix of cell features of each class to predict a closest class, wherein the matrix of cell features has been generated from at least initial training data comprising at least one cell image;
detecting cell pixels from the extracted cell features of the at least one new cell image using the predicted closest class to generate a likelihood map;
extracting individual cells from the at least one cell image by segmenting the individual cells from the likelihood map;
performing a machine annotation on the extracted individual cells from the at least one new cell image to identify cells, non-cell pixels, and/or cell boundaries;
calculating a confidence level for the machine annotation on the extracted individual cells from the at least one new cell image, wherein the calculation of the confidence level comprises:
calculating a confidence score S:
S
=
C
×
∑
i
=
1
M
⁢
⁢

P
i
⁡
(
f
❘
x
i
)
-
P
i
⁡
(
b
❘
x
i
)

M
wherein M is a total number of pixels, Pi(f|xi) is a probability of a pixel xi being classified to a cell pixel while P1(b|x1) is a probability of a pixel xi being classified to a background pixel (not a cell pixel);
and
wherein C is a coefficient and its value is determined by an indication whether the at least one cell image belongs to a new class or an existing class;
and
modifying the machine annotation if the confidence level is below a predetermined threshold.

43 more elements

A non-transitory computer readable medium containing a computer program storing computer readable code for cell annotation, the program being executable by a computer to cause the computer to perform a process comprising:

receiving at least one new cell image for cell detection;
extracting cell features from the at least one new cell image;
comparing the extracted cell features to a matrix of cell features of each class to predict a closest class, wherein the matrix of cell features has been generated from at least initial training data comprising at least one cell image;
detecting cell pixels from the extracted cell features of the at least one new cell image using the predicted closest class to generate a likelihood map;
extracting individual cells from the at least one cell image by segmenting the individual cells from the likelihood map;
performing a machine annotation on the extracted individual cells from the at least one new cell image to identify cells, non-cell pixels, and/or cell boundaries;
calculating a confidence level for the machine annotation on the extracted individual cells from the at least one new cell image, wherein the calculation of the confidence level comprises:
calculating a confidence score S:
S
=
C
×
∑
i
=
1
M
⁢
⁢

P
i
⁡
(
f
❘
x
i
)
-
P
i
⁡
(
b
❘
x
i
)

M
wherein M is a total number of pixels, Pi(f|xi) is a probability of a pixel xi being classified to a cell pixel while Pi(b|xi) is a probability of a pixel xi being classified to a background pixel (not a cell pixel);
and
wherein C is a coefficient and its value is determined by an indication whether the at least one cell image belongs to a new class or an existing class;
and
modifying the machine annotation if the confidence level is below a predetermined threshold.

43 more elements

A system for cell annotation, the system comprising:

at least one processor configured to:
receive at least one new cell image for cell detection;
extract cell features from the at least one new cell image;
compare the extracted cell features to a matrix of cell features of each class to predict a closest class, wherein the matrix of cell features has been generated from at least initial training data comprising at least one cell image;
detect cell pixels from the extracted cell features of the at least one new cell image using the predicted closest class to generate a likelihood map;
extract individual cells from the at least one cell image by segmenting the individual cells from the likelihood map;
perform a machine annotation on the extracted individual cells from the at least one new cell image to identify cells, non-cell pixels, and/or cell boundaries;
calculate a confidence level for the machine annotation on the extracted individual cells from the at least one new cell image, wherein the calculation of the confidence level comprises:
calculating a confidence score S:
S
=
C
×
∑
i
=
1
M
⁢
⁢

P
i
⁡
(
f
❘
x
i
)
-
P
i
⁡
(
b
❘
x
i
)

M
wherein M is a total number of pixels, Pi(f|xi) is a probability of a pixel xi being classified to a cell pixel while Pi(b|xi) is a probability of a pixel xi being classified to a background pixel (not a cell pixel);
and
wherein C is a coefficient and its value is determined by an indication whether the at least one cell image belongs to a new class or an existing class;
and
modify the machine annotation if the confidence level is below a predetermined threshold.

44 more elements

Dependencies are inferred from the claim text.

Facts

Application: 16/306,490Filed 27 Jun 2017 · us provisional application
Priority: 30 Jun 2016
Published: US 2019/0180147 A113 Jun 2019
Granted: 1 Dec 2020
Term ends: Not estimatedThe source records don’t support an estimate.

Classification13 codes

G06N 20/00
Machine learning
G06F 18/217
Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation: validation; Performance…
G06F 18/22
Analysing: matching criteria, e.g. proximity measures
G06F 18/2413
Relating to the classification model, e.g. parametric or non-parametric approaches: based on distances to training or reference patterns