Cell Detection with Star-convex Polygons

By Uwe Schmidt, Martin Weigert, Coleman Broaddus, Eugene W. Myers. International Conference on Medical Image Computing and Computer-Assisted Intervention, 2018

1,640 citations recorded by the source corpus.

Automatic detection and segmentation of cells and nuclei in microscopy images is important for many biological applications. Recent successful learning-based approaches include per-pixel cell segmentation with subsequent pixel grouping, or localization of bounding boxes with subsequent shape refinement. In situations of crowded cells, these can be prone to segmentation errors, such as falsely merging bordering cells or suppressing valid cell instances due to the poor approximation with bounding boxes. To overcome these issues, we propose to localize cell nuclei via star-convex polygons, which are a much better shape representation as compared to bounding boxes and thus do not need shape refinement. To that end, we train a convolutional neural network that predicts for every pixel a polygon for the cell instance at that position. We demonstrate the merits of our approach on two synthetic datasets and one challenging dataset of diverse fluorescence microscopy images.

Subfields: Artificial Intelligence, Biophysics, Computer Vision and Pattern Recognition

View Paper: https://www.semanticscholar.org/paper/82c5d56fb0f4aef16766bdf99c4ad0d81aa86ccf

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Artificial Intelligence

In Biophysics

In Computer Vision and Pattern Recognition

Citations in the corpus

Metrics Reloaded: Recommendations for image analysis validation (2022, neutral): “Despite these problems, the terminology has been adapted by subsequent influential work [49, 76, 82] indicating a wide-spread usage within the community.”
CellProfiler 4: improvements in speed, utility and usability (2021, methodology): “For CellProfiler 3 we previously released a plugin for NucleAIzer [15], and in the future we hope to investigate integrations with other popular models such as Cellpose [16] and Stardist [17].”
TrackMate 7: integrating state-of-the-art segmentation algorithms into tracking pipelines (2022, methodology): “To track MCF10DCIS.com cells labeled with sir-DNA using StarDist and TrackMate, a custom StarDist model was generated using the ZeroCostDL4Mic platform 14,15 .”
Labkit: Labeling and Segmentation Toolkit for Big Image Data (2021, methodology): “Quality metrics scores were calculated as the average precision with threshold 0.5 as deﬁned in StarDist (Schmidt et al., 2018).”
Democratising deep learning for microscopy with ZeroCostDL4Mic (2021, methodology): “These models can then generate masks that can be re ﬁ ned by users (i.e., in Fiji 21 ) before being used for training to obtain a high-performance model 5,7,8,11,23,69 .”
Star-convex Polyhedra for 3D Object Detection and Segmentation in Microscopy (2019, methodology): “Contributions The extension of [21] from 2D to 3D is challenging and our main contribution in this paper.”
PolarMask: Single Shot Instance Segmentation With Polar Representation (2019, methodology): “Polar representation was firstly used in [28] to detect cells in microscopic images, where the problem is much simpler as there are only two categories.”
The ImageJ ecosystem: Open‐source software for image visualization, processing, and analysis (2020, methodology): “Today, CSBDeep has become the basis for many other useful applications.(61,92,97,98) Additionally, the Fijiintegrated Java parts of CSBDeep are now capable (via the use of ImageJ-TensorFlow) to not only run saved deep learning models, but even train such models as Noise2Void and DenoiSeg.”
CellViT: Vision Transformers for Precise Cell Segmentation and Classification (2023, methodology): “Other used libraries include the o ffi cial STARDIST [44], CPP-Net [29] and CellSeg-models implementations [72].”