Evaluation of Deep Learning Strategies for Nucleus Segmentation in Fluorescence Images

By Juan C. Caicedo, Jonathan F Roth, Allen Goodman, Tim Becker, Kyle W. Karhohs, Claire McQuin, Shantanu Singh, Anne E Carpenter. 2018

308 citations recorded by the source corpus.

Identifying nuclei is often a critical first step in analyzing microscopy images of cells, and classical image processing algorithms are most commonly used for this task. Recent developments in deep learning can yield superior accuracy, but typical evaluation metrics for nucleus segmentation do not satisfactorily capture error modes that are relevant in cellular images. Besides, large image data sets with ground truth for evaluation have been limiting. We present an evaluation framework to measure accuracy, types of errors, and computational efficiency; and use it to compare two deep learning strategies (U-Net and DeepCell) alongside a classical approach implemented in CellProfiler. We publicly release a set of 23,165 manually annotated nuclei and source code to reproduce experiments. Our results show that U-Net outperforms both pixel-wise classification networks and classical algorithms. Also, our evaluation framework shows that deep learning improves accuracy and reduces the number of biologically relevant errors by half.

Subfields: Artificial Intelligence, Biophysics, Media Technology, Medical Imaging

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

1. Kevin Smith
2. Anne E Carpenter
3. William Graf
4. Erick Moen
5. Markus W. Covert
6. Takamasa Kudo
7. David Van Valen
8. D. Bannon

In Biophysics

1. Allen Goodman
2. Uwe Schmidt
3. Anne E Carpenter
4. Hanchuan Peng
5. Shantanu Singh
6. Florian Jug
7. Oren Z. Kraus

In Media Technology

1. Uwe Schmidt
2. Anne E Carpenter

In Medical Imaging

1. Anne E Carpenter

Citations in the corpus

• Cell Detection with Star-convex Polygons (2018, background): “While this approach often gives good results, it is problematic for images of very crowded cell nuclei, since only a few mis-classified pixels can cause bordering but distinct cell instances to be fused [3,19].”
• A deep learning framework for nucleus segmentation using image style transfer (2019, methodology): “In addition, our proposed method outperformed all prior published results on hist , a diverse set of histology images and on fluo , a fluorescent image set (BBBC039 9 ​ ) (see Supplementary data for details).”
• Deep learning for cellular image analysis (2019, methodology): “Larger datasets, such as those featured in the 2018 Kaggle Data Science Bowl 103 , render these approaches even more accurate 52 .”