CRISP: correlation-refined image segmentation process
Abstract Background Calcium imaging enables real-time recording of cellular activity across various biological contexts. To assess the activity of individual cells, researchers must segment images into the individual cells. While intensity-based threshold algorithms allow for automatic image segment...
Saved in:
| Main Authors: | , , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
BMC
2025-05-01
|
| Series: | BMC Bioinformatics |
| Subjects: | |
| Online Access: | https://doi.org/10.1186/s12859-025-06150-z |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| _version_ | 1849704787211190272 |
|---|---|
| author | Jennifer K. Briggs Erli Jin Matthew J. Merrins Richard K. P. Benninger |
| author_facet | Jennifer K. Briggs Erli Jin Matthew J. Merrins Richard K. P. Benninger |
| author_sort | Jennifer K. Briggs |
| collection | DOAJ |
| description | Abstract Background Calcium imaging enables real-time recording of cellular activity across various biological contexts. To assess the activity of individual cells, researchers must segment images into the individual cells. While intensity-based threshold algorithms allow for automatic image segmentation in sparsely packed tissues, they perform poorly in densely packed organs such as cardiomyocytes or the pancreatic islet. To study these tissues, investigators typically manually outline the cells based on visual inspection. This manual cell masking introduces potential user error. To address this error, we developed the Correlation-Refined Image Segmentation Process (CRISP). CRISP utilizes interpixel correlations to refine user drawn cell masks (cell mask refinement) or automatically masks cells by identifying the largest circle that captures only pixels within the cell (semi-minor axis identification). Results CRISP cell mask refinement had an area under the receiver operating curve of 0.835, indicating good model performance on the training data set. CRISP had 77% accuracy when testing on a separate data set, which came from a different mouse model imaged with a different microscope than the training data set. CRISP cell mask refinement significantly improved the accuracy of functional network analysis compared to non-CRISP refined cell masks. CRISP automated semi-minor axis identification had an area under the receiver operating curve under the curve of 0.989, indicating strong model performance. Conclusions Inaccurate cell masking can result in inaccurate scientific interpretations of calcium images. Utilizing interpixel correlations, we developed two transparent algorithms that can be used for image segmentation in densely packed tissues. These algorithms may allow for more accurate and reproducible cell masking. |
| format | Article |
| id | doaj-art-ef5d419a465747fc9da48844efce3fe7 |
| institution | DOAJ |
| issn | 1471-2105 |
| language | English |
| publishDate | 2025-05-01 |
| publisher | BMC |
| record_format | Article |
| series | BMC Bioinformatics |
| spelling | doaj-art-ef5d419a465747fc9da48844efce3fe72025-08-20T03:16:39ZengBMCBMC Bioinformatics1471-21052025-05-0126111410.1186/s12859-025-06150-zCRISP: correlation-refined image segmentation processJennifer K. Briggs0Erli Jin1Matthew J. Merrins2Richard K. P. Benninger3Department of Bioengineering, University of Colorado Anschutz Medical CampusDivision of Endocrinology, Diabetes & Metabolism, Department of Medicine, University of Wisconsin-MadisonDivision of Endocrinology, Diabetes & Metabolism, Department of Medicine, University of Wisconsin-MadisonDepartment of Bioengineering, University of Colorado Anschutz Medical CampusAbstract Background Calcium imaging enables real-time recording of cellular activity across various biological contexts. To assess the activity of individual cells, researchers must segment images into the individual cells. While intensity-based threshold algorithms allow for automatic image segmentation in sparsely packed tissues, they perform poorly in densely packed organs such as cardiomyocytes or the pancreatic islet. To study these tissues, investigators typically manually outline the cells based on visual inspection. This manual cell masking introduces potential user error. To address this error, we developed the Correlation-Refined Image Segmentation Process (CRISP). CRISP utilizes interpixel correlations to refine user drawn cell masks (cell mask refinement) or automatically masks cells by identifying the largest circle that captures only pixels within the cell (semi-minor axis identification). Results CRISP cell mask refinement had an area under the receiver operating curve of 0.835, indicating good model performance on the training data set. CRISP had 77% accuracy when testing on a separate data set, which came from a different mouse model imaged with a different microscope than the training data set. CRISP cell mask refinement significantly improved the accuracy of functional network analysis compared to non-CRISP refined cell masks. CRISP automated semi-minor axis identification had an area under the receiver operating curve under the curve of 0.989, indicating strong model performance. Conclusions Inaccurate cell masking can result in inaccurate scientific interpretations of calcium images. Utilizing interpixel correlations, we developed two transparent algorithms that can be used for image segmentation in densely packed tissues. These algorithms may allow for more accurate and reproducible cell masking.https://doi.org/10.1186/s12859-025-06150-zAutomated image segmentationCalcium imagingCell maskingCorrelation |
| spellingShingle | Jennifer K. Briggs Erli Jin Matthew J. Merrins Richard K. P. Benninger CRISP: correlation-refined image segmentation process BMC Bioinformatics Automated image segmentation Calcium imaging Cell masking Correlation |
| title | CRISP: correlation-refined image segmentation process |
| title_full | CRISP: correlation-refined image segmentation process |
| title_fullStr | CRISP: correlation-refined image segmentation process |
| title_full_unstemmed | CRISP: correlation-refined image segmentation process |
| title_short | CRISP: correlation-refined image segmentation process |
| title_sort | crisp correlation refined image segmentation process |
| topic | Automated image segmentation Calcium imaging Cell masking Correlation |
| url | https://doi.org/10.1186/s12859-025-06150-z |
| work_keys_str_mv | AT jenniferkbriggs crispcorrelationrefinedimagesegmentationprocess AT erlijin crispcorrelationrefinedimagesegmentationprocess AT matthewjmerrins crispcorrelationrefinedimagesegmentationprocess AT richardkpbenninger crispcorrelationrefinedimagesegmentationprocess |