This is a preprint, a preliminary version of a manuscript that has not completed peer review at a journal. Research Square does not conduct peer review prior to posting preprints. The posting of a preprint on this server should not be interpreted as an endorsement of its validity or suitability for dissemination as established information or for guiding clinical practice.
Article
Cecilia Lindskog
Uppsala University
Corresponding Author
ORCiD: https://orcid.org/0000-0001-5611-1015
License:
This work is licensed under a CC BY 4.0 License. Read Full License
Immunohistochemistry (IHC) provides the basis for cell type-specific localization of protein expression patterns in human tissues. Manual annotation of complex IHC images is however expensive and may lead to errors or inter-observer variability. Artificial Intelligence holds much promise for efficient and accurate pattern recognition, but confidence in prediction needs to be addressed. To present a reliable model for annotation of IHC images, we developed a semi-automated framework for multi-label classification of 7,848 human testis samples stained with IHC, and manually annotated in situ protein expression in eight different cell types. The dataset was used as a basis for training and testing a proposed Hybrid Bayesian Neural Network. By combining the deep learning model with a novel uncertainty metric, the average diagnostic performance improved from 86.9% to 96.3%. The streamlined workflow has important implications for accurate large-scale efforts mapping the human cell type-specific proteome in health and disease.
Keywords
immunohistochemistry, human cell type-specific proteome, human testis
Figure 1
Figure 2
Figure 3
Figure 4
Figure 5
Figure 6
This is a list of supplementary files associated with this preprint. Click to download.
- SupplementaryFigure1.pdf
Model accuracy and percentage of discarded images based on GTL threshold. The plots show the accuracy and percentage of discarded images for different values of GTL threshold when applied to the HBNet for each cell type. Notice how the plots illustrate the tradeoff between accuracy (orange) and discard rate (blue), where high GTL thresholds result in better accuracy but at the expense of discarding large numbers of images, whereas low thresholds result in the opposite. For most cell types there is a large step increase for the first increments of threshold (between 0.0 and 0.3) followed by far smaller increments (for thresholds over 0.3). This makes choosing a threshold cutoff easy. However, for round/early spermatids the increment is more gradual and constant, resulting in a less clear cut decision point.
- SupplementaryTable1.xlsx
List of images included in the test dataset. Information on Ensembl ID, antibody ID, gene name, gene description, URLs to Human Protein Atlas JPEG image, as well as manual annotation, model prediction, GTL Score and manual evaluation of staining intensity and subcellular localization for each of the eight analyzed cell types.
Loading...
Badges
Prescreen
This manuscript passed our Prescreen assessment
Complete author information
Appropriate declaration statements
Potential risks to human health
Prescreen
History
CURRENT STATUS: Posted
Version 1
Posted 05 Aug, 2020
No community comments so far
Metrics
Comments: 0
PDF Downloads: ...
HTML Views: ...
Learn more about our company.
This is a preprint. It has not completed peer review.
Article
Cecilia Lindskog
Uppsala University
Corresponding Author
ORCiD: https://orcid.org/0000-0001-5611-1015
Badges
Prescreen
This manuscript passed our Prescreen assessment
Complete author information
Appropriate declaration statements
Potential risks to human health
Prescreen
History
CURRENT STATUS: Posted
Version 1
Posted 05 Aug, 2020
No community comments so far
License:
This work is licensed under a CC BY 4.0 License. Read Full License
Immunohistochemistry (IHC) provides the basis for cell type-specific localization of protein expression patterns in human tissues. Manual annotation of complex IHC images is however expensive and may lead to errors or inter-observer variability. Artificial Intelligence holds much promise for efficient and accurate pattern recognition, but confidence in prediction needs to be addressed. To present a reliable model for annotation of IHC images, we developed a semi-automated framework for multi-label classification of 7,848 human testis samples stained with IHC, and manually annotated in situ protein expression in eight different cell types. The dataset was used as a basis for training and testing a proposed Hybrid Bayesian Neural Network. By combining the deep learning model with a novel uncertainty metric, the average diagnostic performance improved from 86.9% to 96.3%. The streamlined workflow has important implications for accurate large-scale efforts mapping the human cell type-specific proteome in health and disease.
Figure 1
Figure 2
Figure 3
Figure 4
Figure 5
Figure 6
Loading...