# ACDC: Automated Cell Detection and Counting for Time-Lapse Fluorescence Microscopy

^{1}

^{2}

^{3}

^{4}

^{5}

^{6}

^{7}

^{8}

^{9}

^{10}

^{11}

^{*}

^{†}

^{‡}

## Abstract

**:**

## Featured Application

**Novel method for Automated Cell Detection and Counting (ACDC) in time-lapse fluorescence microscopy.**

## Abstract

## 1. Introduction

- ACDC is designed and developed to cope with the analysis of stacks of time-lapse microscopy images in real-time;
- ACDC does not require any training phase, and represents a reliable solution even without the availability of large-scale annotated datasets.

## 2. Materials and Methods

#### 2.1. Fluorescence Microscopy Imaging Data

#### 2.1.1. Vanderbilt University Dataset

#### 2.1.2. 2018 Data Science Bowl

#### 2.2. Acdc: A Method for the Automatic Cell Detection and Counting

#### 2.2.1. Pre-Processing

- Application of bilateral filtering that allows for denoising the image $\mathcal{I}$ while preserving the edges by means of a non-linear combination of nearby image values [38]. This noise-reducing smoothing filter combines gray levels (colors) according to both a geometric closeness function c and a radiometric (photometric) similarity function s. This combination is used to strengthen near values with respect to distant values in both spatial and intensity domains. This simple yet effective strategy allows for contrast enhancement [49]. Bilateral filter has been shown to work properly in fluorescence imaging even preserving the directional information, such as in the case of the F-actin filaments [50]. This denoising technique was effectively applied to biological electron microscopy [51], as well as to cell detection [52], revealing better performance—compared to low-pass filtering—in noise reduction without removing the structural features conveyed by strong edges. The most commonly used version of bilateral filtering is the shift-invariant Gaussian filtering, wherein both the closeness function c and the similarity function s are Gaussian functions of the Euclidean distance between their arguments [38]. With more details, c is radially symmetric: $c(\mathbf{p},\mathbf{q})={e}^{-\frac{1}{2}{\left(\frac{\left|\right|\mathbf{p}-\mathbf{q}\left|\right|}{{\sigma}_{s}}\right)}^{2}}$. Consistently, the similarity function s can be defined as: $s(\mathbf{p},\mathbf{q})={e}^{-\frac{1}{2}{\left(\frac{\left|\right|I\left(\mathbf{p}\right)-I\left(\mathbf{q}\right||}{{\sigma}_{c}})\right)}^{2}}$. In ACDC we set ${\sigma}_{c}=1$ and ${\sigma}_{s}={\sigma}_{\mathrm{global}}$ (where ${\sigma}_{\mathrm{global}}$ is the the standard deviation of the input image $\mathcal{I}$) for the standard deviation of the Gaussian functions c and s, respectively. This smart denoising approach allows us to keep the edge sharpness while reducing the noise of the processed image, so avoiding cell region under-estimation.
- Application of top-hat transform for background correction with a binary circular structuring element (radius: 21 pixels) on the smoothed image. This operation accounts for non-uniform illumination artifacts, by extracting the nuclei from the background. The white top-hat transform is the difference between the input image I and the opening of $\mathcal{I}$ with a gray-scale structuring element b: ${\mathcal{T}}_{\mathrm{w}}=\mathcal{I}-\mathcal{I}\circ b$ [53].

#### 2.2.2. Nucleus Seed Selection

- A thresholding technique has to be first applied to detect the cell regions. Both global and local thresholding techniques aim at separating foreground objects of interest from the background in an image, considering differences in pixel intensities [54]. Global thresholding determines a single threshold for all pixels and works well if the histogram of the input image contains well-separated peaks corresponding to the desired foreground objects and background [55]. Local adaptive thresholding techniques estimate the threshold locally over sub-regions of the entire image, by considering only a user-defined window with a specific size and exploiting local image properties to calculate a variable threshold [53,54]. These algorithms find the threshold by locally examining the intensity values of the neighborhood of each pixel according to image intensity statistics. To avoid unwanted pixels in the thresholded image, mainly due to small noisy hyper-intense regions caused by non-uniform illumination, we apply the Otsu global thresholding method [55] instead of local adaptive thresholding based on the mean value in a neighborhood [56]. Moreover, global threshold techniques are significantly faster than local adaptive strategies.
- Hole filling is applied to remove possible holes in the detected nuclei due to small hypo-intense regions included in the nuclei regions.
- Morphological opening (using a disk with 1-pixel radius as a structuring element) is used to remove loosely connected-components, such as in the case of almost overlapping cells.
- Unwanted areas are removed according to the connected-components size. In particular, the detected candidate regions with areas smaller than 40 pixels are removed to refine the achieved segmentation results by robustly avoiding false positives.
- Morphological closing (using a 2-pixel radius circular structuring element) is applied to smooth the boundaries of the detected nuclei and avoid the under-estimation of the detected nuclei regions.
- The approximate Euclidean distance transform (EDT) from the binary mask, achieved by applying the Otsu algorithm and refined by using the previous 3 steps, is used to obtain the matrix of distances of each pixel to the background by exploiting the ${\ell}_{2}$ Euclidean distance [57] (with a $5\times 5$ pixel mask for a more accurate distance estimation). This algorithm calculates the distance to the closest background pixel for each pixel of the source image. Let $\mathcal{G}$ be a regular grid and $f:\mathcal{G}\to \mathbb{R}$ an arbitrary function on the grid, called a sampled function [58]. We define the distance transform ${\mathcal{D}}_{f}:\mathcal{G}\to \mathbb{R}$ of f as:$${\mathcal{D}}_{f}\left(\mathbf{p}\right)=\underset{\mathbf{q}\in \mathcal{G}}{min}\left(d(\mathbf{p},\mathbf{q})+f\left(\mathbf{q}\right)\right),$$$${\mathcal{D}}_{\mathcal{P}}=\underset{\mathbf{q}\in \mathcal{P}}{min}\left(d(\mathbf{p},\mathbf{q})+\mathrm{\U0001d7d9}\left(\mathbf{q}\right)\right),$$$$\mathrm{\U0001d7d9}\left(\mathbf{q}\right)=\left\{\begin{array}{cc}0,\hfill & \mathrm{if}\phantom{\rule{4.pt}{0ex}}\mathbf{q}\in \mathcal{P}\hfill \\ \infty ,\hfill & \mathrm{otherwise}\hfill \end{array}\right.$$
- Regional maxima computation allows for estimating foreground peaks on the normalized distance map. Regional maxima are connected-components of pixels with a constant intensity value, whose external boundary pixels have all a lower intensity value [42]. The resulting binary mask contains pixels that are set to 1 for identifying regional maxima, while all other pixels are set to 0. A $5\times 5$ pixel square was employed as structuring element.
- Morphological dilation (using a 3-pixel radius circular structuring element) is applied to the foreground peaks previously detected for better defining the foreground regions and merging neighboring local minima into a single seed point. The segmentation results on Figure 5a,b are shown in Figure 6a,b, respectively. The detail in Figure 5a shows that ACDC is highly specific to cell nuclei detection, discarding non-cell regions related to acquisition artifacts.

#### 2.2.3. Cell Nuclei Segmentation Using the Watershed Transform

#### 2.2.4. Implementation Details

`mpi4py`, which provides bindings of the Message Passing Interface (MPI) specifications for Python to leverage multi-core and many-core resources [69]. The distributed strategy used to accelerate ACDC is similar to that employed in [70,71,72], where the Parent allocates the resources and orchestrates the workers, which run ACDC to analyze the assigned images. This distributed version of ACDC is $3.7\times $ faster than the sequential version by exploiting 6 cores of a CPU Intel Core E5-2650 v4 (clock $2.2$ GHz).

#### 2.3. Segmentation Evaluation Metrics

## 3. Results

#### 3.1. ACDC Performance

#### 3.2. Comparison with Other Cell Imaging Tools and Segmentation Methods

## 4. Discussion and Conclusions

## Author Contributions

## Funding

## Acknowledgments

## Conflicts of Interest

## References

- Kanade, T.; Yin, Z.; Bise, R.; Huh, S.; Eom, S.; Sandbothe, M.F.; Chen, M. Cell image analysis: Algorithms, system and applications. In Proceedings of the IEEE Workshop on Applications of Computer Vision (WACV), Kona, HI, USA, 5–7 January 2011; pp. 374–381. [Google Scholar] [CrossRef] [Green Version]
- Orth, J.D.; Kohler, R.H.; Foijer, F.; Sorger, P.K.; Weissleder, R.; Mitchison, T.J. Analysis of mitosis and antimitotic drug responses in tumors by in vivo microscopy and single-cell pharmacodynamics. Cancer Res.
**2011**, 71, 4608–4616. [Google Scholar] [CrossRef] [PubMed] [Green Version] - Manandhar, S.; Bouthemy, P.; Welf, E.; Danuser, G.; Roudot, P.; Kervrann, C. 3D flow field estimation and assessment for live cell fluorescence microscopy. Bioinformatics
**2020**, 36, 1317–1325. [Google Scholar] [CrossRef] [PubMed] [Green Version] - Peng, H. Bioimage informatics: A new area of engineering biology. Bioinformatics
**2008**, 24, 1827–1836. [Google Scholar] [CrossRef] [PubMed] [Green Version] - Meijering, E.; Carpenter, A.E.; Peng, H.; Hamprecht, F.A.; Olivo-Marin, J.C. Imagining the future of bioimage analysis. Nat. Biotechnol.
**2016**, 34, 1250. [Google Scholar] [CrossRef] [PubMed] - Peng, H.; Bateman, A.; Valencia, A.; Wren, J.D. Bioimage informatics: a new category in Bioinformatics. Bioinformatics
**2012**, 28, 1057. [Google Scholar] [CrossRef] [Green Version] - Meijering, E. Cell segmentation: 50 years down the road [life sciences]. IEEE Signal Process. Mag.
**2012**, 29, 140–145. [Google Scholar] [CrossRef] - Schneider, C.A.; Rasband, W.S.; Eliceiri, K.W. NIH Image to ImageJ: 25 years of image analysis. Nat. Methods
**2012**, 9, 671. [Google Scholar] [CrossRef] - Schindelin, J.; Arganda-Carreras, I.; Frise, E.; Kaynig, V.; Longair, M.; Pietzsch, T.; Preibisch, S.; Rueden, C.; Saalfeld, S.; Schmid, B.; et al. Fiji: An open-source platform for biological-image analysis. Nat. Methods
**2012**, 9, 676. [Google Scholar] [CrossRef] [Green Version] - Carpenter, A.E.; Jones, T.R.; Lamprecht, M.R.; Clarke, C.; Kang, I.H.; Friman, O. CellProfiler: Image analysis software for identifying and quantifying cell phenotypes. Genome Biol.
**2006**, 7, R100. [Google Scholar] [CrossRef] [Green Version] - Dao, D.; Fraser, A.N.; Hung, J.; Ljosa, V.; Singh, S.; Carpenter, A.E. CellProfiler Analyst: Interactive data exploration, analysis and classification of large biological image sets. Bioinformatics
**2016**, 32, 3210–3212. [Google Scholar] [CrossRef] - Wählby, C.; Sintorn, I.M.; Erlandsson, F.; Borgefors, G.; Bengtsson, E. Combining intensity, edge and shape information for 2D and 3D segmentation of cell nuclei in tissue sections. J. Microsc.
**2004**, 215, 67–76. [Google Scholar] [CrossRef] [PubMed] - Kaliman, S.; Jayachandran, C.; Rehfeldt, F.; Smith, A.S. Limits of Applicability of the Voronoi Tessellation Determined by Centers of Cell Nuclei to Epithelium Morphology. Front. Physiol.
**2016**, 7, 551. [Google Scholar] [CrossRef] [PubMed] [Green Version] - Honda, H. Description of cellular patterns by Dirichlet domains: The two-dimensional case. J. Theor. Biol.
**1978**, 72, 523–543. [Google Scholar] [CrossRef] - Kostrykin, L.; Schnörr, C.; Rohr, K. Globally optimal segmentation of cell nuclei in fluorescence microscopy images using shape and intensity information. Med. Image Anal.
**2019**, 58, 101536. [Google Scholar] [CrossRef] - Gamarra, M.; Zurek, E.; Escalante, H.J.; Hurtado, L.; San-Juan-Vergara, H. Split and merge watershed: A two-step method for cell segmentation in fluorescence microscopy images. Biomed. Signal Process. Control
**2019**, 53, 101575. [Google Scholar] [CrossRef] - Angermueller, C.; Pärnamaa, T.; Parts, L.; Stegle, O. Deep learning for computational biology. Mol. Syst. Biol.
**2016**, 12, 878. [Google Scholar] [CrossRef] - Berg, S.; Kutra, D.; Kroeger, T.; Straehle, C.N.; Kausler, B.X.; Haubold, C.; Schiegg, M.; Ales, J.; Beier, T.; Rudy, M.; et al. ilastik: Interactive machine learning for (bio) image analysis. Nat. Methods
**2019**, 16, 1226–1232. [Google Scholar] [CrossRef] - Held, M.; Schmitz, M.H.; Fischer, B.; Walter, T.; Neumann, B.; Olma, M.H.; Peter, M.; Ellenberg, J.; Gerlich, D.W. CellCognition: time-resolved phenotype annotation in high-throughput live cell imaging. Nat. Methods
**2010**, 7, 747. [Google Scholar] [CrossRef] [Green Version] - Ciresan, D.; Giusti, A.; Gambardella, L.M.; Schmidhuber, J. Deep Neural Networks Segment Neuronal Membranes in Electron Microscopy Images; Advances in Neural Information Processing Systems (NIPS): Lake Tahoe, NV, USA, 2012; pp. 2843–2851. [Google Scholar]
- Rosati, R.; Romeo, L.; Silvestri, S.; Marcheggiani, F.; Tiano, L.; Frontoni, E. Faster R-CNN approach for detection and quantification of DNA damage in comet assay images. Comput. Biol. Med.
**2020**, 103912. [Google Scholar] [CrossRef] - Sadanandan, S.K.; Ranefall, P.; Le Guyader, S.; Wählby, C. Automated Training of Deep Convolutional Neural Networks for Cell Segmentation. Sci. Rep.
**2017**, 7, 7860. [Google Scholar] [CrossRef] [Green Version] - Hiramatsu, Y.; Hotta, K.; Imanishi, A.; Matsuda, M.; Terai, K.; Liu, D.; Zhang, D.; Song, Y.; Zhang, C.; Huang, H.; et al. Cell Image Segmentation by Integrating Multiple CNNs. In Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition (CVPR) Workshops, Salt Lake City, UT, USA, 18–22 June 2018; pp. 2205–2211. [Google Scholar]
- Ren, S.; He, K.; Girshick, R.; Sun, J. Faster R-CNN: Towards real-time object detection with region proposal networks. In Proceedings of the Advances in Neural Information Processing Systems (NIPS), Montrea, QC, Canada, 7–12 December 2015; pp. 91–99. [Google Scholar]
- Redmon, J.; Divvala, S.; Girshick, R.; Farhadi, A. You only look once: Unified, real-time object detection. In Proceedings of the Conference on Computer Vision and Pattern Recognition (CVPR), Las Vegas, NV, USA, 27–30 June 2016; pp. 779–788. [Google Scholar] [CrossRef] [Green Version]
- Alam, M.M.; Islam, M.T. Machine learning approach of automatic identification and counting of blood cells. Healthc. Technol. Lett.
**2019**, 6, 103–108. [Google Scholar] [CrossRef] - Han, C.; Kitamura, Y.; Kudo, A.; Ichinose, A.; Rundo, L.; Furukawa, Y.; Umemoto, K.; Li, Y.; Nakayama, H. Synthesizing diverse lung nodules wherever massively: 3D multi-conditional GAN-based CT image augmentation for object detection. In Proceedings of the International Conference on 3D Vision (3DV), Quebec, QC, Canada, 16–19 September 2019; pp. 729–737. [Google Scholar] [CrossRef] [Green Version]
- Bayramoglu, N.; Heikkilä, J. Transfer learning for cell nuclei classification in histopathology images. In Proceedings of the European Conference on Computer Vision (ECCV) Workshops; Springer: Berlin/Heidelberg, Germany, 2016; Volume 9915, pp. 532–539. [Google Scholar] [CrossRef]
- Apicella, A.; Isgrò, F.; Prevete, R. A simple and efficient architecture for trainable activation functions. Neurocomputing
**2019**, 370, 1–15. [Google Scholar] [CrossRef] [Green Version] - Pelt, D.M.; Sethian, J.A. A mixed-scale dense convolutional neural network for image analysis. Proc. Natl. Acad. Sci. USA
**2018**, 115, 254–259. [Google Scholar] [CrossRef] [PubMed] [Green Version] - Osokin, A.; Chessel, A.; Carazo Salas, R.E.; Vaggi, F. GANs for biological image synthesis. In Proceedings of the IEEE International Conference on Computer Vision (ICCV), Venice, Italy, 22–29 October 2017; pp. 2233–2242. [Google Scholar] [CrossRef] [Green Version]
- Han, C.; Rundo, L.; Araki, R.; Furukawa, Y.; Mauri, G.; Nakayama, H.; Hayashi, H. Infinite brain MR images: PGGAN-based data augmentation for tumor detection. In Neural Approaches to Dynamics of Signal Exchanges; Smart Innovation, Systems and Technologies; Springer: Berlin/Heidelberg, Germany, 2019; Volume 151, pp. 291–303. [Google Scholar] [CrossRef] [Green Version]
- Lo Castro, D.; Tegolo, D.; Valenti, C. A visual framework to create photorealistic retinal vessels for diagnosis purposes. J. Biomed. Inform.
**2020**, 103490. [Google Scholar] [CrossRef] [PubMed] - Kraus, O.Z.; Ba, J.L.; Frey, B.J. Classifying and segmenting microscopy images with deep multiple instance learning. Bioinformatics
**2016**, 32, i52–i59. [Google Scholar] [CrossRef] [PubMed] - Militello, C.; Rundo, L.; Minafra, L.; Cammarata, F.P.; Calvaruso, M.; Conti, V.; Russo, G. MF2C3: Multi-Feature Fuzzy Clustering to Enhance Cell Colony Detection in Automated Clonogenic Assay Evaluation. Symmetry
**2020**, 12, 773. [Google Scholar] [CrossRef] - Meyer, C.T.; Wooten, D.J.; Paudel, B.B.; Bauer, J.; Hardeman, K.N.; Westover, D.; Lovly, C.M.; Harris, L.A.; Tyson, D.R.; Quaranta, V. Quantifying drug combination synergy along potency and efficacy axes. Cell Syst.
**2019**, 8, 97–108. [Google Scholar] [CrossRef] [Green Version] - Caicedo, J.C.; Goodman, A.; Karhohs, K.W.; Cimini, B.A.; Ackerman, J.; Haghighi, M.; Heng, C.; Becker, T.; Doan, M.; McQuin, C.; et al. Nucleus segmentation across imaging experiments: The 2018 Data Science Bowl. Nat. Methods
**2019**, 16, 1247–1253. [Google Scholar] [CrossRef] - Tomasi, C.; Manduchi, R. Bilateral filtering for gray and color images. In Proceedings of the Sixth International Conference on Computer Vision (ICCV), Bombay, India, 4–7 January 1998; pp. 839–846. [Google Scholar] [CrossRef]
- Soille, P.J.; Ansoult, M.M. Automated basin delineation from digital elevation models using mathematical morphology. Signal Process.
**1990**, 20, 171–182. [Google Scholar] [CrossRef] - Vincent, L.; Soille, P. Watersheds in digital spaces: an efficient algorithm based on immersion simulations. IEEE Trans. Pattern Anal. Mach. Intell.
**1991**, 13, 583–598. [Google Scholar] [CrossRef] [Green Version] - Beucher, S.; Meyer, F. The morphological approach to segmentation: The watershed transformation. In Mathematical Morphology in Image Processing; Marcel Dekker Inc.: New York, NY, USA, 1993; Volume 34, pp. 433–481. [Google Scholar]
- Soille, P. Morphological Image Analysis: Principles and Applications, 2nd ed.; Springer Science & Business Media: Secaucus, NJ, USA, 2004. [Google Scholar] [CrossRef]
- Tyson, D.R.; Garbett, S.P.; Frick, P.L.; Quaranta, V. Fractional proliferation: A method to deconvolve cell population dynamics from single-cell data. Nat. Methods
**2012**, 9, 923. [Google Scholar] [CrossRef] [PubMed] [Green Version] - Harris, L.A.; Frick, P.L.; Garbett, S.P.; Hardeman, K.N.; Paudel, B.B.; Lopez, C.F.; Quaranta, V.; Tyson, D.R. An unbiased metric of antiproliferative drug effect in vitro. Nat. Methods
**2016**, 13, 497–500. [Google Scholar] [CrossRef] [PubMed] - Sakaue-Sawano, A.; Kurokawa, H.; Morimura, T.; Hanyu, A.; Hama, H.; Osawa, H.; Kashiwagi, S.; Fukami, K.; Miyata, T.; Miyoshi, H.; et al. Visualizing spatiotemporal dynamics of multicellular cell-cycle progression. Cell
**2008**, 132, 487–498. [Google Scholar] [CrossRef] [PubMed] [Green Version] - Kaggle. 2018 Data Science Bowl. 2018. Available online: https://www.kaggle.com/c/data-science-bowl-2018 (accessed on 14 December 2019).
- Georgescu, W.; Wikswo, J.P.; Quaranta, V. CellAnimation: An open source MATLAB framework for microscopy assays. Bioinformatics
**2011**, 28, 138–139. [Google Scholar] [CrossRef] [PubMed] [Green Version] - Sansone, M.; Zeni, O.; Esposito, G. Automated segmentation of comet assay images using Gaussian filtering and fuzzy clustering. Med. Biol. Eng. Comput.
**2012**, 50, 523–532. [Google Scholar] [CrossRef] [PubMed] - Schettini, R.; Gasparini, F.; Corchs, S.; Marini, F.; Capra, A.; Castorina, A. Contrast image correction method. J. Electron. Imaging
**2010**, 19, 023005. [Google Scholar] [CrossRef] - Venkatesh, M.; Mohan, K.; Seelamantula, C.S. Directional bilateral filters for smoothing fluorescence microscopy images. AIP Advances
**2015**, 5, 084805. [Google Scholar] [CrossRef] - Jiang, W.; Baker, M.L.; Wu, Q.; Bajaj, C.; Chiu, W. Applications of a bilateral denoising filter in biological electron microscopy. J. Struct. Biol.
**2003**, 144, 114–122. [Google Scholar] [CrossRef] - Li, K.; Miller, E.D.; Chen, M.; Kanade, T.; Weiss, L.E.; Campbell, P.G. Computer vision tracking of stemness. In Proceedings of the 5th IEEE International Symposium on Biomedical Imaging: From Nano to Macro (ISBI), Paris, France, 14–17 May 2008; pp. 847–850. [Google Scholar] [CrossRef] [Green Version]
- Gonzalez, R.; Woods, R. Digital Image Processing, 3rd ed.; Prentice Hall Press: Upper Saddle River, NJ, USA, 2002. [Google Scholar]
- Jain, A.K. Fundamentals of Digital Image Processing, 1st ed.; Prentice Hall Press: Upper Saddle River, NJ, USA, 2002. [Google Scholar]
- Otsu, N. A threshold selection method from gray-level histograms. IEEE Trans. Syst. Man Cybern.
**1975**, 11, 23–27. [Google Scholar] [CrossRef] [Green Version] - Militello, C.; Rundo, L.; Conti, V.; Minafra, L.; Cammarata, F.P.; Mauri, G.; Gilardi, M.C.; Porcino, N. Area-based cell colony surviving fraction evaluation: A novel fully automatic approach using general-purpose acquisition hardware. Comput. Biol. Med.
**2017**, 89, 454–465. [Google Scholar] [CrossRef] - Borgefors, G. Distance transformations in digital images. Comput. Vis. Graph. Image Process.
**1986**, 34, 344–371. [Google Scholar] [CrossRef] - Felzenszwalb, P.F.; Huttenlocher, D.P. Distance transforms of sampled functions. Theory Comput.
**2012**, 8, 415–428. [Google Scholar] [CrossRef] - Salvi, M.; Morbiducci, U.; Amadeo, F.; Santoro, R.; Angelini, F.; Chimenti, I.; Massai, D.; Messina, E.; Giacomello, A.; Pesce, M.; et al. Automated segmentation of fluorescence microscopy images for 3D cell detection in human-derived cardiospheres. Sci. Rep.
**2019**, 9, 6644. [Google Scholar] [CrossRef] [PubMed] [Green Version] - Grau, V.; Mewes, A.; Alcaniz, M.; Kikinis, R.; Warfield, S.K. Improved watershed transform for medical image segmentation using prior information. IEEE Trans. Med. Imaging
**2004**, 23, 447–458. [Google Scholar] [CrossRef] [PubMed] - Suzuki, K.; Horiba, I.; Sugie, N. Linear-time connected-component labeling based on sequential local operations. Comput. Vis. Image Underst.
**2003**, 89, 1–23. [Google Scholar] [CrossRef] - Meyer, F. Topographic distance and watershed lines. Signal Process.
**1994**, 38, 113–125. [Google Scholar] [CrossRef] - Najman, L.; Couprie, M.; Bertrand, G. Watersheds, mosaics, and the emergence paradigm. Discrete Appl. Math.
**2005**, 147, 301–324. [Google Scholar] [CrossRef] [Green Version] - Van der Walt, S.; Schönberger, J.L.; Nunez-Iglesias, J.; Boulogne, F.; Warner, J.D.; Yager, N.; Gouillart, E.; Yu, T.; the scikit-image contributors. scikit-image: Image processing in Python. PeerJ
**2014**, 2, e453. [Google Scholar] [CrossRef] - Coelho, L.P. Mahotas: Open source software for scriptable computer vision. J. Open Res. Softw.
**2013**, 1, e3. [Google Scholar] [CrossRef] - Rundo, L.; Militello, C.; Vitabile, S.; Casarino, C.; Russo, G.; Midiri, M.; Gilardi, M.C. Combining split-and-merge and multi-seed region growing algorithms for uterine fibroid segmentation in MRgFUS treatments. Med. Biol. Eng. Comput.
**2016**, 54, 1071–1084. [Google Scholar] [CrossRef] - Celery Project. Celery Distributed Task Queue. 2018. Available online: http://www.celeryproject.org/ (accessed on 14 December 2019).
- Pivotal Software, Inc. RabbitMQ. 2018. Available online: http://www.rabbitmq.com/ (accessed on 14 December 2019).
- Dalcín, L.; Paz, R.; Storti, M. MPI for Python. J. Parallel Distrib. Comput.
**2005**, 65, 1108–1115. [Google Scholar] [CrossRef] - Rundo, L.; Tangherloni, A.; Cazzaniga, P.; Nobile, M.S.; Russo, G.; Gilardi, M.C.; Vitabile, S.; Mauri, G.; Besozzi, D.; Militello, C. A novel framework for MR image segmentation and quantification by using MedGA. Comput. Methods Programs Biomed.
**2019**, 176, 159–172. [Google Scholar] [CrossRef] [PubMed] - Tangherloni, A.; Spolaor, S.; Rundo, L.; Nobile, M.S.; Cazzaniga, P.; Mauri, G.; Liò, P.; Merelli, I.; Besozzi, D. GenHap: A novel computational method based on genetic algorithms for haplotype assembly. BMC Bioinform.
**2019**, 20, 172. [Google Scholar] [CrossRef] [PubMed] [Green Version] - Tangherloni, A.; Rundo, L.; Spolaor, S.; Cazzaniga, P.; Nobile, M.S. GPU-powered multi-swarm parameter estimation of biological systems: A master-slave approach. In Proceedings of the 26th Euromicro International Conference on Parallel, Distributed and Network-based Processing, Cambridge, UK, 21–23 March 2018; pp. 698–705. [Google Scholar]
- Bland, J.M.; Altman, D.G. Measuring agreement in method comparison studies. Stat. Methods Med. Res.
**1999**, 8, 135–160. [Google Scholar] [CrossRef] [PubMed] - Wilcoxon, F. Individual comparisons by ranking methods. Biometrics Bull.
**1945**, 1, 80–83, 196–202. [Google Scholar] [CrossRef] - Wiliem, A.; Wong, Y.; Sanderson, C.; Hobson, P.; Chen, S.; Lovell, B.C. Classification of human epithelial type 2 cell indirect immunofluoresence images via codebook based descriptors. In Proceedings of the IEEE Workshop on Applications of Computer Vision (WACV), Tampa, FL, USA, 15–17 January 2013; pp. 95–102. [Google Scholar] [CrossRef] [Green Version]
- Coelho, L.P.; Shariff, A.; Murphy, R.F. Nuclear segmentation in microscope cell images: A hand-segmented dataset and comparison of algorithms. In Proceedings of the IEEE International Symposium on Biomedical Imaging (ISBI): From Nano to Macro, Boston, MA, USA, 28 June–1 July 2009; pp. 518–521. [Google Scholar] [CrossRef] [Green Version]
- Osuna, E.G.; Hua, J.; Bateman, N.W.; Zhao, T.; Berget, P.B.; Murphy, R.F. Large-scale automated analysis of location patterns in randomly tagged 3T3 cells. Ann. Biomed. Eng.
**2007**, 35, 1081–1087. [Google Scholar] [CrossRef] [Green Version] - Kraus, O.Z.; Grys, B.T.; Ba, J.; Chong, Y.; Frey, B.J.; Boone, C.; Andrews, B.J. Automated analysis of high-content microscopy data with deep learning. Mol. Syst. Biol.
**2017**, 13, 924. [Google Scholar] [CrossRef] [PubMed] - Win, K.; Choomchuay, S.; Hamamoto, K.; Raveesunthornkiat, M. Detection and Classification of Overlapping Cell Nuclei in Cytology Effusion Images Using a Double-Strategy Random Forest. Appl. Sci.
**2018**, 8, 1608. [Google Scholar] [CrossRef] [Green Version] - Salvi, M.; Cerrato, V.; Buffo, A.; Molinari, F. Automated segmentation of brain cells for clonal analyses in fluorescence microscopy images. J. Neurosci. Methods
**2019**, 325, 108348. [Google Scholar] [CrossRef]

**Figure 1.**(

**a**,

**b**) Examples of the analyzed microscopy fluorescence images provided by the Department of Biochemistry of the VU. The images were displayed by automatically adjusting the brightness and contrast according to an histogram-based procedure.

**Figure 3.**Boxplots depicting the distribution for both the analyzed datasets in terms of: (

**a**) total number of cells, and (

**b**) coverage of the cell nuclei regions.

**Figure 4.**Flow diagram of the ACDC pipeline. The gray, black and light-blue data blocks denote gray-scale images, binary masks and information extracted from the images, respectively. The three macro-blocks represent the three main processing phases, namely: pre-processing, seed selection, and watershed-based segmentation.

**Figure 8.**Comparison of the gold standard cell nuclei segmentation (magenta contour in the left images) against the automated result obtained by ACDC (segmented nuclei over-imposed onto the original fluorescence images with alpha-blending in the right images): (

**a**) VU dataset, where the orange arrows denote errors in the split of clustered cell nuclei; (

**b**) DSB dataset, where the green arrows denote groups of cells that were erroneously delineated as unique connected-components in the gold standard and the blue dashed boxes represent spurious speckles that are not detected by ACDC.

**Figure 9.**Scatter plots depicting ACDC results compared to the gold standard in terms of cell nuclei counting in the case of: (

**a**) time-lapse fluorescence images from the VU dataset; (

**b**) small fluorescent nuclei images from the DSB dataset. The equality line through the origin is drawn as a dashed line.

**Figure 10.**Bland-Altman plots of the cell counting measurements achieved by ACDC versus the gold standard for the (

**a**) VU and (

**b**) DSB datasets. Solid horizontal and dashed lines denote the mean and $\pm 1.96$ standard deviation values, respectively.

**Figure 11.**Regplots showing the scatter plots obtained by considering the number of manually detected cells (y-axis) and the number of cells automatically detected (x-axis) with ACDC (left), ImageJ with Gaussian filter (center), and ImageJ without Gaussian filter (right), along with the fitted regression model (regression line and the $95\%$ confidence interval for that regression). Plots (

**a**–

**c**) report the results obtained on the VU dataset, while plots (

**d**–

**f**) are obtained from the DSB dataset.

**Figure 12.**Examples of cell nuclei segmented by using ACDC and the implemented ImageJ pipelines with and without Gaussian filtering. (

**a**,

**b**) present images taken from the VU dataset, with different visual characteristics.

**Table 1.**Evaluation metrics on cell counting and segmentation achieved by ACDC (with and without bilateral filtering) on the analyzed time-lapse microscopy VU and 2018 DSB datasets, comprising 46 and 301 images, respectively. The results for the DSC and IoU metrics, as well as the execution time measurements, are expressed as mean value ± standard deviation.

Method | Dataset | Pearson Coeff. (p-Value) | DSC (%) | IoU (%) | Exec. Time (s) |
---|---|---|---|---|---|

ACDC (without bilateral filter) | VU | $\rho =0.99$$(p=2.5\times {10}^{-74})$ | $75.86\pm 5.98$ | $61.45\pm 7.47$ | $3.98\pm 0.10$ |

ACDC (with bilateral filter) | VU | $\rho =0.99$$(p=6.6\times {10}^{-74})$ | $76.84\pm 6.71$ | $62.84\pm 8.58$ | $7.49\pm 0.30$ |

ACDC (without bilateral filter) | DSB | $\rho =0.96$$(p=1.1\times {10}^{-175})$ | $87.34\pm 6.89$ | $77.97\pm 9.49$ | $0.07\pm 0.05$ |

ACDC (with bilateral filter) | DSB | $\rho =0.96$$(p=2.6\times {10}^{-169})$ | $88.64\pm 7.41$ | $80.37\pm 10.58$ | $0.12\pm 0.09$ |

**Table 2.**Comparison of ACDC and the tested ImageJ pipelines in terms of cell nuclei counting and segmentation. The results are expressed as mean value ± standard deviation.

Method | Dataset | Pearson Coeff. (p-Value) | DSC (%) | IoU (%) |
---|---|---|---|---|

ACDC | VU | $\rho =0.99$$(p=6.6\times {10}^{-74})$ | $76.84\pm 6.71$ | $62.84\pm 8.58$ |

ImageJ (with Gaussian filter) | VU | $\rho =0.99$$(p=9.5\times {10}^{-75})$ | $74.97\pm 6.17$ | $60.32\pm 7.62$ |

ImageJ (without Gaussian filter) | VU | $\rho =0.99$$(p=5.3\times {10}^{-44})$ | $74.43\pm 6.20$ | $59.62\pm 7.61$ |

ACDC | DSB | $\rho =0.96$$(p=2.6\times {10}^{-169})$ | $88.64\pm 7.41$ | $80.37\pm 10.58$ |

ImageJ (with Gaussian filter) | DSB | $\rho =0.97$$(p=5.7\times {10}^{-205})$ | $86.50\pm 6.86$ | $76.78\pm 9.66$ |

ImageJ (without Gaussian filter) | DSB | $\rho =0.97$$(p=3.4\times {10}^{-207})$ | $86.68\pm 7.75$ | $77.23\pm 10.86$ |

**Table 3.**Comparison of ACDC, CellProfiler, MC-Watershed, and SM-Watershed in terms of cell nuclei segmentation on the SNPHEp-2 dataset.

Method | DSC (%) | IoU (%) |
---|---|---|

ACDC | $80.92$ | $67.96$ |

CellProfiler | $78.12$ | $64.10$ |

MC-Watershed | $86.54$ | $76.27$ |

SM-Watershed | $83.10$ | $71.09$ |

© 2020 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).

## Share and Cite

**MDPI and ACS Style**

Rundo, L.; Tangherloni, A.; Tyson, D.R.; Betta, R.; Militello, C.; Spolaor, S.; Nobile, M.S.; Besozzi, D.; Lubbock, A.L.R.; Quaranta, V.;
et al. ACDC: Automated Cell Detection and Counting for Time-Lapse Fluorescence Microscopy. *Appl. Sci.* **2020**, *10*, 6187.
https://doi.org/10.3390/app10186187

**AMA Style**

Rundo L, Tangherloni A, Tyson DR, Betta R, Militello C, Spolaor S, Nobile MS, Besozzi D, Lubbock ALR, Quaranta V,
et al. ACDC: Automated Cell Detection and Counting for Time-Lapse Fluorescence Microscopy. *Applied Sciences*. 2020; 10(18):6187.
https://doi.org/10.3390/app10186187

**Chicago/Turabian Style**

Rundo, Leonardo, Andrea Tangherloni, Darren R. Tyson, Riccardo Betta, Carmelo Militello, Simone Spolaor, Marco S. Nobile, Daniela Besozzi, Alexander L. R. Lubbock, Vito Quaranta,
and et al. 2020. "ACDC: Automated Cell Detection and Counting for Time-Lapse Fluorescence Microscopy" *Applied Sciences* 10, no. 18: 6187.
https://doi.org/10.3390/app10186187