Method for robust analysis of biological activity in microscopy images
View Patent ↗An object analysis method performs object segmentation to generate segmentation confidence map and uses the segmentation results to generate robust object features. The robust object features are combined to create robust FOV summary features, robust sample summary features and robust assay summary features.
1. An object segmentation confidence mapping method for analysis of biological activity comprising the steps of:
a) Input an image;
b) Perform segmentation decision using the input image having the segmentation decision result output;
c) Input a threshold;
d) Perform difference operation using the segmentation decision result and the threshold having the difference result output;
e) Perform confidence mapping using the difference result and a confidence function having the segmentation confidence output wherein the confidence function is determined using at least one training image to determine segmentation accuracy for different distance values wherein the empirically determined segmentation accuracy function is normalized or scaled as the confidence function.
2. A robust object segmentation method for analysis of biological activity comprising the steps of:
a) Input an image;
b) Perform segmentation confidence mapping using the input image having the segmentation confidence map output;
c) Perform thresholding using the object segmentation confidence map having the high confidence object mask output;
d) Perform another thresholding using the object segmentation confidence map having the low confidence object mask output;
e) Display object segmentation result using the high confidence object mask;
f) Perform robust object feature measurement using the low confidence object mask having the robust object features output.
3. The method of claim 2 wherein the object segmentation confidence mapping method further comprises the steps of:
a) Perform segmentation decision using the input image having the segmentation decision result output;
b) Input a threshold;
c) Perform difference operation using the segmentation decision result and the threshold having the difference result output;
d) Perform confidence mapping using the difference result and a confidence function having the segmentation confidence output wherein the confidence function is determined using at least one training image to determine segmentation accuracy for different distance values wherein the empirically determined segmentation accuracy function is normalized or scaled as the confidence function.
4. A FOV regulated feature extraction method for biological activity comprising the steps of:
a) Input a plurality of object feature results;
b) Perform control object selection using the plurality of object feature results having the control objects output;
c) Perform FOV regulation feature extraction using the control objects having the FOV regulation features output wherein the FOV regulation feature extraction calculates using the following formula:
F
FOV_r
=
F
-
θ
R
FOV
1
γ
+
(
1
-
γ
)
R
FOV
2
d) Perform object feature FOV regulation using the plurality of object feature results and the FOV regulation features having the FOV regulated object features output wherein the regulated features is generated using center or dispersion selected control FOVs.