↓ Skip to main content

Fusion of encoder-decoder deep networks improves delineation of multiple nuclear phenotypes

Overview of attention for article published in BMC Bioinformatics, August 2018
Altmetric Badge

About this Attention Score

  • Above-average Attention Score compared to outputs of the same age (52nd percentile)
  • Above-average Attention Score compared to outputs of the same age and source (54th percentile)

Mentioned by

twitter
6 X users

Citations

dimensions_citation
16 Dimensions

Readers on

mendeley
40 Mendeley
You are seeing a free-to-access but limited selection of the activity Altmetric has collected about this research output. Click here to find out more.
Title
Fusion of encoder-decoder deep networks improves delineation of multiple nuclear phenotypes
Published in
BMC Bioinformatics, August 2018
DOI 10.1186/s12859-018-2285-0
Pubmed ID
Authors

Mina Khoshdeli, Garrett Winkelmaier, Bahram Parvin

Abstract

Nuclear segmentation is an important step for profiling aberrant regions of histology sections. If nuclear segmentation can be resolved, then new biomarkers of nuclear phenotypes and their organization can be predicted for the application of precision medicine. However, segmentation is a complex problem as a result of variations in nuclear geometry (e.g., size, shape), nuclear type (e.g., epithelial, fibroblast), nuclear phenotypes (e.g., vesicular, aneuploidy), and overlapping nuclei. The problem is further complicated as a result of variations in sample preparation (e.g., fixation, staining). Our hypothesis is that (i) deep learning techniques can learn complex phenotypic signatures that rise in tumor sections, and (ii) fusion of different representations (e.g., regions, boundaries) contributes to improved nuclear segmentation. We have demonstrated that training of deep encoder-decoder convolutional networks overcomes complexities associated with multiple nuclear phenotypes, where we evaluate alternative architecture of deep learning for an improved performance against the simplicity of the design. In addition, improved nuclear segmentation is achieved by color decomposition and combining region- and boundary-based features through a fusion network. The trained models have been evaluated against approximately 19,000 manually annotated nuclei, and object-level Precision, Recall, F1-score and Standard Error are reported with the best F1-score being 0.91. Raw training images, annotated images, processed images, and source codes are released as a part of the Additional file 1. There are two intrinsic barriers in nuclear segmentation to H&E stained images, which correspond to the diversity of nuclear phenotypes and perceptual boundaries between adjacent cells. We demonstrate that (i) the encoder-decoder architecture can learn complex phenotypes that include the vesicular type; (ii) delineation of overlapping nuclei is enhanced by fusion of region- and edge-based networks; (iii) fusion of ENets produces an improved result over the fusion of UNets; and (iv) fusion of networks is better than multitask learning. We suggest that our protocol enables processing a large cohort of whole slide images for applications in precision medicine.

X Demographics

X Demographics

The data shown below were collected from the profiles of 6 X users who shared this research output. Click here to find out more about how the information was compiled.
Mendeley readers

Mendeley readers

The data shown below were compiled from readership statistics for 40 Mendeley readers of this research output. Click here to see the associated Mendeley record.

Geographical breakdown

Country Count As %
Unknown 40 100%

Demographic breakdown

Readers by professional status Count As %
Student > Master 11 28%
Student > Ph. D. Student 5 13%
Other 3 8%
Student > Doctoral Student 3 8%
Student > Bachelor 2 5%
Other 2 5%
Unknown 14 35%
Readers by discipline Count As %
Computer Science 9 23%
Medicine and Dentistry 4 10%
Engineering 3 8%
Biochemistry, Genetics and Molecular Biology 2 5%
Social Sciences 2 5%
Other 3 8%
Unknown 17 43%
Attention Score in Context

Attention Score in Context

This research output has an Altmetric Attention Score of 3. This is our high-level measure of the quality and quantity of online attention that it has received. This Attention Score, as well as the ranking and number of research outputs shown below, was calculated when the research output was last mentioned on 25 August 2018.
All research outputs
#13,045,234
of 23,344,526 outputs
Outputs from BMC Bioinformatics
#3,691
of 7,387 outputs
Outputs of similar age
#155,291
of 331,401 outputs
Outputs of similar age from BMC Bioinformatics
#41
of 98 outputs
Altmetric has tracked 23,344,526 research outputs across all sources so far. This one is in the 43rd percentile – i.e., 43% of other outputs scored the same or lower than it.
So far Altmetric has tracked 7,387 research outputs from this source. They typically receive a little more attention than average, with a mean Attention Score of 5.5. This one is in the 48th percentile – i.e., 48% of its peers scored the same or lower than it.
Older research outputs will score higher simply because they've had more time to accumulate mentions. To account for age we can compare this Altmetric Attention Score to the 331,401 tracked outputs that were published within six weeks on either side of this one in any source. This one has gotten more attention than average, scoring higher than 52% of its contemporaries.
We're also able to compare this research output to 98 others from the same source and published within six weeks on either side of this one. This one has gotten more attention than average, scoring higher than 54% of its contemporaries.