DART: A design-aware microfluidic chip paradigm for real-time live-cell image analysis

Dietrich Kohlheyer; Hanno Scharr; Johannes Seiffarth; Katharina N\"oh; Lukas Scholtes; Matthias Pesch

arxiv: 2606.18523 · v1 · pith:XEEQGAFUnew · submitted 2026-06-16 · 🧬 q-bio.QM · cs.CV

DART: A design-aware microfluidic chip paradigm for real-time live-cell image analysis

Johannes Seiffarth , Matthias Pesch , Lukas Scholtes , Dietrich Kohlheyer , Hanno Scharr , Katharina N\"oh This is my paper

Pith reviewed 2026-06-26 21:24 UTC · model grok-4.3

classification 🧬 q-bio.QM cs.CV

keywords microfluidic chipslive-cell imagingfiducial markersdeep learningimage registrationreal-time analysissingle-cell dataCAD alignment

0 comments

The pith

DART aligns CAD blueprints to physical microfluidic chips via fiducial markers and deep learning to enable throughput-independent real-time image analysis.

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The paper presents the DART paradigm to overcome scaling limits in high-throughput live-cell imaging where manual or semi-automated RoI localization and structure removal slow down analysis. It embeds fiducial markers in the chip design and trains a deep learning model to detect them, thereby registering the CAD layout directly to each acquired image. This registration permits automatic extraction of every RoI and removal of surrounding microfluidic features without depending on the total number of regions. Validation on a chip containing 1164 RoIs of eight different geometries shows localization in five minutes and per-image processing in under 1.1 seconds. The resulting pipeline supports fully automated downstream tasks such as cell segmentation.

Core claim

DART establishes alignment through embedded fiducial markers and deep-learning-based marker detection, enabling throughput-independent localization of all RoIs and fully automated image processing across diverse RoI geometries and chip layouts.

What carries the argument

The DART paradigm, which registers the CAD blueprint to the physical chip using fiducial markers detected by a deep learning model.

If this is right

All RoIs on a chip with 1164 locations are localized in five minutes regardless of count.
Microfluidic structures are removed from raw images in 40 ms each.
Fully automated cell segmentation and analysis complete in under 1.1 s per image.
The pipeline supports closed-loop and outcome-driven smart microscopy experiments.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

The same registration step could be combined with real-time feedback to adjust imaging parameters or fluidic conditions during an experiment.
Marker-based alignment may transfer to microfluidic devices used for chemical assays or organ-on-chip models beyond live-cell work.
Automated removal of device features could reduce batch-to-batch variability when the same analysis code is applied across different fabrication runs.

Load-bearing premise

Fiducial markers stay reliably detectable and produce alignment accurate enough to excise microfluidic structures without artifacts or missed cells on every layout and imaging condition.

What would settle it

A new chip layout or imaging condition where the marker detector fails on more than a few percent of markers, producing visible misalignment or cell loss after structure removal.

Figures

Figures reproduced from arXiv: 2606.18523 by Dietrich Kohlheyer, Hanno Scharr, Johannes Seiffarth, Katharina N\"oh, Lukas Scholtes, Matthias Pesch.

**Figure 1.** Figure 1: Overview of the DART microfluidic cultivation platform. High-throughput live-cell imaging is limited by (A) time-consuming manual RoI localization, (B) semi-automated image processing, and (C) offline image analysis that delays quantitative readouts such as single-cell or population development. (D) The DART paradigm addresses these limitations by aligning the CAD blueprint with the physical microfluidic c… view at source ↗

**Figure 2.** Figure 2: CAD layout of the SAK chip and its RoI architectures. (A) Overview of the full chip, including the inlets and outlets of the four independently supplied media channels. (B) Closeup of the grid layout within a single channel. RoIs are arranged in rows and columns; the first two digits of the unique ID encode the row position and the last two digits encode the column position. Each RoI is further equipped w… view at source ↗

**Figure 3.** Figure 3: Aligning CAD blueprint and microscope stage coordinate systems. (A) The operator visits three RoIs and records an image, the corresponding RoI ID, and the microscope stage position. DART detects the fiducial markers, determines the RoI centers, and computes an affine transformation from the corresponding coordinate pairs in blueprint and stage coordinates. This transformation maps the CAD blueprint (B) on… view at source ↗

**Figure 4.** Figure 4: DART’s design-aware fine alignment and per-image processing pipeline. (A) Fivestep image-processing workflow for two exemplary RoI designs comprising (1) fiducial marker detection, (2) marker-pair matching, (3) rotation and translation into the blueprint coordinate system, (4) blueprint-derived masking and cropping of microfluidic structures, and (5) cell segmentation. (B) Mean processing time of each pip… view at source ↗

**Figure 5.** Figure 5: Live-cell validation of the DART pipeline on the SAK chip. (A) Representative time-lapse excerpts of C. glutamicum populations cultivated in different SAK chamber geometries with complex interior structures. (B) Temporal development of the TSCA, derived from automated DART masking and cell segmentation (solid) and fitted logistic growth models (dashed) demonstrating automated extraction of quantitative g… view at source ↗

read the original abstract

High-throughput microfluidic live-cell imaging generates rich single-cell data. Yet semi-automated procedures for locating regions of interest (RoIs), each containing one cell population, and removing surrounding microfluidic structures from recorded images, scale with the number of RoIs. This prevents real-time image analysis and delays time-to-insight by hours to days. We introduce the Design-Aware and Real-Time capable (DART) paradigm for microfluidic cultivation chips, which aligns the CAD blueprint with the physical chip and thereby enables throughput-independent localization of all RoIs and fully automated image processing across diverse RoI geometries and chip layouts. DART establishes this alignment through embedded fiducial markers and deep-learning-based marker detection. We validate DART using the Swiss Army Knife chip, which combines eight structurally distinct RoI designs across 1164 RoI locations. DART localizes all RoIs in five minutes, removes microfluidic structures from raw microscopy images in 40 ms, and performs fully automated image analysis, including cell segmentation, in under 1.1 s per image. Together, these capabilities establish DART as an end-to-end hardware-software paradigm with real-time-capable analysis that paves the way toward closed-loop and outcome-driven smart microscopy.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Desk Editor's Note private letter to a colleague

DART shows a working fiducial-plus-DL alignment trick that decouples RoI localization from throughput on one complex chip, but supplies no error metrics or robustness tests to back the accuracy claim.

read the letter

The paper's core move is to embed fiducial markers in the microfluidic chip so a deep-learning detector can register the CAD layout to the physical device. Once aligned, every RoI can be found and the surrounding structures masked without the usual per-RoI manual work. They demonstrate this on the Swiss Army Knife chip that packs eight different RoI geometries into 1164 locations.

The reported timings are the clearest evidence of progress: five minutes to localize everything, 40 ms to strip the microfluidic background, and under 1.1 s for full automated analysis including segmentation. Those numbers suggest the pipeline could actually run in real time and support closed-loop experiments.

The limitation is straightforward. All numbers come from a single chip design under unspecified imaging conditions. No marker-detection RMSE, no post-alignment overlay error in pixels or microns, no failure rate when focus, illumination, or magnification changes, and no comparison against standard registration methods. Without those figures it is impossible to judge whether the alignment is accurate enough to avoid cell loss or residual artifacts across the claimed range of layouts.

The work is aimed at labs already running high-throughput microfluidic live-cell experiments who are bottlenecked by post-acquisition processing. Anyone building similar chips would find the fiducial strategy and the end-to-end timing data worth seeing.

It should go to peer review. The concrete performance on a non-trivial device is enough to justify referee time, even though the missing quantitative validation will be the main point reviewers need to press.

Referee Report

2 major / 2 minor

Summary. The manuscript introduces the DART paradigm for microfluidic live-cell imaging chips. It uses embedded fiducial markers and deep-learning-based detection to align CAD blueprints with physical chips, enabling throughput-independent RoI localization and fully automated image processing (including structure removal and cell segmentation) that does not scale with the number of RoIs. Validation is reported on the Swiss Army Knife chip containing eight distinct RoI designs across 1164 locations, with claims of 5-minute localization, 40 ms structure removal, and <1.1 s per-image analysis.

Significance. If the alignment step proves robust, DART would remove a key scalability bottleneck in high-throughput microfluidic imaging by making analysis time independent of RoI count and layout diversity. This could support closed-loop smart microscopy applications. The approach is presented as a hardware-software co-design without reliance on fitted parameters or self-referential quantities.

major comments (2)

[Validation / Results] Validation section (Swiss Army Knife chip results): The central claim that fiducial-based alignment enables artifact-free microfluidic structure removal and cell-preserving analysis rests on the unquantified assumption of sufficient DL marker detection reliability and sub-cellular alignment precision. No marker localization RMSE, post-alignment overlay error (in pixels or microns), detection precision/recall, or failure rates under changed focus/illumination/magnification are reported, despite testing 1164 RoIs on one chip design.
[Abstract / Methods] Abstract and methods description of alignment pipeline: The claim of 'throughput-independent' and 'fully automated' processing across 'diverse RoI geometries and chip layouts' is load-bearing for the paradigm's novelty, yet the manuscript provides no cross-layout or cross-condition statistics (e.g., success rate on held-out chip designs or under imaging variations) to support generalization beyond the single tested multi-design chip.

minor comments (2)

[Abstract] The abstract reports timing numbers (5 min, 40 ms, <1.1 s) without specifying the hardware platform or whether these include DL inference time on CPU/GPU.
[Validation] No baseline comparisons (e.g., manual RoI annotation time or conventional template-matching alignment) are mentioned to contextualize the reported speedups.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for the constructive feedback highlighting areas where quantitative support and generalization evidence can be strengthened. We address each major comment below and indicate planned revisions.

read point-by-point responses

Referee: [Validation / Results] Validation section (Swiss Army Knife chip results): The central claim that fiducial-based alignment enables artifact-free microfluidic structure removal and cell-preserving analysis rests on the unquantified assumption of sufficient DL marker detection reliability and sub-cellular alignment precision. No marker localization RMSE, post-alignment overlay error (in pixels or microns), detection precision/recall, or failure rates under changed focus/illumination/magnification are reported, despite testing 1164 RoIs on one chip design.

Authors: We agree that explicit quantitative metrics on marker detection reliability and alignment precision are not reported in the current manuscript and would strengthen the validation claims. The successful localization and processing across all 1164 RoIs on the Swiss Army Knife chip provides indirect evidence of reliability, but we will add a dedicated analysis subsection reporting detection precision/recall, localization RMSE, estimated post-alignment overlay error, and any available data on performance under focus or illumination variations from the existing experiments. revision: yes
Referee: [Abstract / Methods] Abstract and methods description of alignment pipeline: The claim of 'throughput-independent' and 'fully automated' processing across 'diverse RoI geometries and chip layouts' is load-bearing for the paradigm's novelty, yet the manuscript provides no cross-layout or cross-condition statistics (e.g., success rate on held-out chip designs or under imaging variations) to support generalization beyond the single tested multi-design chip.

Authors: The Swiss Army Knife chip validation incorporates eight structurally distinct RoI designs, demonstrating performance across varied geometries within one layout. We acknowledge the absence of tests on held-out chip designs or multiple independent layouts, which limits claims of broad generalization. We will revise the abstract and methods sections to more precisely scope the claims to the tested multi-design chip while retaining the throughput-independence result, and add discussion of the diversity covered by the eight RoI types. revision: partial

Circularity Check

0 steps flagged

No circularity: hardware-software paradigm with no derivations or fitted predictions

full rationale

The paper presents DART as a paradigm using embedded fiducial markers and DL-based detection to align CAD blueprints with physical chips, enabling automated RoI localization and image processing. Validation is reported on the Swiss Army Knife chip (1164 RoIs localized in 5 min, processing in 40 ms). No equations, parameter fitting to data subsets, self-citations as load-bearing premises, or renamings of known results appear in the abstract or described claims. The central claims rest on empirical system performance rather than any self-referential reduction by construction.

Axiom & Free-Parameter Ledger

0 free parameters · 0 axioms · 0 invented entities

Abstract-only review provides no explicit free parameters, axioms, or invented entities; DART is described as a paradigm rather than a derivation relying on new postulates.

pith-pipeline@v0.9.1-grok · 5768 in / 1084 out tokens · 27008 ms · 2026-06-26T21:24:15.081117+00:00 · methodology

discussion (0)

Reference graph

Works this paper leans on

36 extracted references · 32 canonical work pages · 1 internal anchor

[1]

Wang, P ., Robert, L., Pelletier, J., Dang, W.L., Taddei, F ., Wright, A., and Jun, S. (2010). Robust Growth of Escherichia coli. Current Biology20, 1099–1103. doi:10.1016/j.cub. 2010.04.045

work page doi:10.1016/j.cub 2010
[2]

Gr ¨unberger, A., Paczia, N., Probst, C., Schendzielorz, G., Eggeling, L., Noack, S., Wiechert, W., and Kohlheyer, D. (2012). A disposable picolitre bioreactor for cultivation and investiga- tion of industrially relevant bacteria on the single cell level. Lab on a Chip12, 2060. doi: 10.1039/c2lc40156h

work page doi:10.1039/c2lc40156h 2012
[5]

Kaiser, M., Jug, F ., Julou, T., Deshpande, S., Pfohl, T., Silander, O.K., Myers, G., and van Nimwegen, E. (2018). Monitoring single-cell gene regulation under dynamically controllable conditions with integrated microfluidics and software. Nature Communications9, 212. doi: 10.1038/s41467-017-02505-0

work page doi:10.1038/s41467-017-02505-0 2018
[6]

Gr ¨unberger, A., Probst, C., Helfrich, S., Nanda, A., Stute, B., Wiechert, W., von Lieres, E., N¨oh, K., Frunzke, J., and Kohlheyer, D. (2015). Spatiotemporal microbial single-cell analysis using a high-throughput microfluidics cultivation platform. Cytometry Part A87, 1101–1115. doi:10.1002/cyto.a.22779

work page doi:10.1002/cyto.a.22779 2015
[7]

Helfrich, S., Pfeifer, E., Kr ¨amer, C., Sachs, C.C., Wiechert, W., Kohlheyer, D., N¨oh, K., and Frunzke, J. (2015). Live cell imaging of SOS and prophage dynamics in isogenic bacterial populations. Molecular Microbiology98, 636–650. doi:10.1111/mmi.13147

work page doi:10.1111/mmi.13147 2015
[8]

Sachs, C.C., Gr ¨unberger, A., Helfrich, S., Probst, C., Wiechert, W., Kohlheyer, D., and N¨oh, K. (2016). Image-based single cell profiling: High-throughput processing of mother machine experiments. PLOS ONE11, e0163453. doi:10.1371/journal.pone.0163453

work page doi:10.1371/journal.pone.0163453 2016
[9]

Bl ¨obaum, L., T¨auber, S., and Gr¨unberger, A. (2023). Protocol to perform dynamic microflu- idic single-cell cultivation of C. glutamicum. STAR Protocols4, 102436. doi:10.1016/j. xpro.2023.102436

work page doi:10.1016/j 2023
[10]

O’Connor, O.M., Alnahhas, R.N., Lugagne, J.B., and Dunlop, M.J. (2022). DeLTA 2.0: A deep learning pipeline for quantifying single-cell spatial and temporal dynamics. PLOS Computational Biology18, e1009797. doi:10.1371/journal.pcbi.1009797. 19

work page doi:10.1371/journal.pcbi.1009797 2022
[11]

Prangemeier, T., Wildner, C., Franc ¸ani, A.O., Reich, C., and Koeppl, H. (2022). Y east cell segmentation in microstructured environments with deep learning. Biosystems211, 104557. doi:10.1016/j.biosystems.2021.104557

work page doi:10.1016/j.biosystems.2021.104557 2022
[12]

Zhou, S., Chen, B., Fu, E.S., and Y an, H. (2023). Computer vision meets microfluidics: a label-free method for high-throughput cell analysis. Microsystems & Nanoengineering9. doi:10.1038/s41378-023-00562-8

work page doi:10.1038/s41378-023-00562-8 2023
[13]

Cutler, K.J., Stringer, C., Lo, T.W., Rappez, L., Stroustrup, N., Brook Peterson, S., Wiggins, P .A., and Mougous, J.D. (2022). Omnipose: a high-precision morphology- independent solution for bacterial cell segmentation. Nature Methods pp. 1–11. doi: 10.1038/s41592-022-01639-4

work page doi:10.1038/s41592-022-01639-4 2022
[14]

Stringer, C., Wang, T., Michaelos, M., and Pachitariu, M. (2020). Cellpose: a gener- alist algorithm for cellular segmentation. Nature Methods18, 100–106. doi:10.1038/ s41592-020-01018-x

2020
[15]

Pachitariu, M., Rariden, M., and Stringer, C. (2025). Cellpose-SAM: superhuman general- ization for cellular segmentation. bioRxiv. doi:10.1101/2025.04.28.651001

work page doi:10.1101/2025.04.28.651001 2025
[16]

Marks, U., Israel, R., Van Valen, D. et al. (2025). A foundation model for cellular segmenta- tion. Nature Methods. doi:10.1038/s41592-025-02879-w

work page doi:10.1038/s41592-025-02879-w 2025
[17]

Archit, A., Freckmann, L., Nair, S., Khalid, N., Hilt, P ., Rajashekar, V., Freitag, M., Teuber, C., Buckley, G., von Haaren, S., Gupta, S., Dengel, A., Ahmed, S., and Pape, C. (2025). Segment anything for microscopy. Nature Methods. doi:10.1038/s41592-024-02580-4

work page doi:10.1038/s41592-024-02580-4 2025
[18]

Kasahara, K., Seiffarth, J., Stute, B., von Lieres, E., Drepper, T., N¨oh, K., and Kohlheyer, D. (2025). Unveiling microbial single-cell growth dynamics under rapid periodic oxygen oscil- lations. Lab on a Chip25, 2234–2246. doi:10.1039/D5LC00065C

work page doi:10.1039/d5lc00065c 2025
[19]

Witting, L., Seiffarth, J., Stute, B., Schulze, T., Hofer, J.M., N ¨oh, K., Eisenhut, M., Weber, A.P .M., von Lieres, E., and Kohlheyer, D. (2025). A microfluidic system for the cultivation of cyanobacteria with precise light intensity and CO2 control: enabling growth data acquisition at single-cell resolution. Lab on a Chip25, 319–329. doi:10.1039/D4LC00567H

work page doi:10.1039/d4lc00567h 2025
[20]

Dal Co, A., van Vliet, S., Kiviet, D.J., Schlegel, S., and Ackermann, M. (2020). Short-range interactions govern the dynamics and functions of microbial communities. Nature Ecology & Evolution4, 366–375. doi:10.1038/s41559-019-1080-2

work page doi:10.1038/s41559-019-1080-2 2020
[21]

Thiermann, R., Sandler, M., Ahir, G., Sauls, J.T., Schroeder, J., Brown, S., Le Treut, G., Si, F ., Li, D., Wang, J.D., and Jun, S. (2024). Tools and methods for high-throughput single-cell imaging with the mother machine. eLife12, RP88463. doi:10.7554/eLife.88463. 20

work page doi:10.7554/elife.88463 2024
[22]

Lugagne, J.B., Blassick, C.M., and Dunlop, M.J. (2024). Deep model predictive control of gene expression in thousands of single cells. Nature Communications15. doi:10.1038/ s41467-024-46361-1

2024
[23]

Merrin, J. (2019). Frontiers in Microfluidics, a Teaching Resource Review. Bioengineering 6, 109. doi:10.3390/bioengineering6040109

work page doi:10.3390/bioengineering6040109 2019
[24]

Long, Z., Nugent, E., Javer, A., Cicuta, P ., Sclavi, B., Lagomarsino, M.C., and Dorfman, K.D. (2013). Microfluidic chemostat for measuring single cell dynamics in bacteria. Lab on a Chip13, 947–954. doi:10.1039/C2LC41196B

work page doi:10.1039/c2lc41196b 2013
[25]

Shi, Y ., Tabet, J.S., Milkie, D.E., Daugird, T.A., Y ang, C.Q., Ritter, A.T., Giovannucci, A., and Legant, W.R. (2024). Smart lattice light-sheet microscopy for imaging rare and complex cellular events. Nature Methods. doi:10.1038/s41592-023-02126-0

work page doi:10.1038/s41592-023-02126-0 2024
[26]

Waithe, D., Brown, J.M., Reglinski, K., Diez-Sevilla, I., Roberts, D., and Eggeling, C. (2020). Object detection networks and augmented reality for cellular detection in fluorescence mi- croscopy. Journal of Cell Biology219. doi:10.1083/jcb.201903166

work page doi:10.1083/jcb.201903166 2020
[27]

Al-Hamadani, M.N.A., Poroszlay, R., Szeman-Nagy, G., Hajdu, A., Hadjidemetriou, S., Ferrarini, L., and Harangi, B. (2025). Improving cell detection and tracking in microscopy images using YOLO and an enhanced DeepSORT algorithm. Sensors25, 4361. doi: 10.3390/s25144361

work page doi:10.3390/s25144361 2025
[28]

Jocher, G., Qiu, J., and Chaurasia, A. (2023). Ultralytics YOLO. GitHub. URL:https: //github.com/ultralytics/ultralytics. doi:10.5281/zenodo.7347926

work page doi:10.5281/zenodo.7347926 2023
[29]

Chiron, L., Le Bec, M., Cordier, C., Pouzet, S., Milunov, D., Banderas, A., Di Meglio, J.M., Sorre, B., and Hersen, P . (2022). CyberSco.Py an open-source software for event-based, conditional microscopy. Scientific Reports12, 11579. doi:10.1038/s41598-022-15207-5

work page doi:10.1038/s41598-022-15207-5 2022
[30]

Passmore, J.B., Rates, A., Schr ¨oder, J., van Laarhoven, M.T.P ., Hellebrekers, V.J.W., van Hoef, H.G., Geurts, A.J.M., van Straaten, W., Nijenhuis, W., Berger, F ., Smith, C.S., Smal, I., and Kapitein, L.C. (2025). Closed-loop optogenetic control of cell biology enables outcome- driven microscopy. Nature Communications17, 1087. doi:10.1038/s41467-025-67848-5

work page doi:10.1038/s41467-025-67848-5 2025
[31]

Gruenberger, A., Probst, C., Heyer, A., Wiechert, W., Frunzke, J., and Kohlheyer, D. (2013). Microfluidic picoliter bioreactor for microbial single-cell analysis: Fabrication, system setup, and operation. Journal of Visualized Experiments. URL:https://app.jove.com/t/50560. doi:10.3791/50560

work page doi:10.3791/50560 2013
[32]

Ravi, N., Gabeur, V., Hu, Y .T., Hu, R., Ryali, C., Ma, T., Khedr, H., R ¨adle, R., Rolland, C., Gustafson, L., Mintun, E., Pan, J., Alwala, K.V., Carion, N., Wu, C.Y ., Girshick, R., Doll´ar, P ., and Feichtenhofer, C. (2024). SAM 2: Segment Anything in Images and Videos. arXiv. doi: 10.48550/arXiv.2408.00714. 21

work page internal anchor Pith review Pith/arXiv arXiv doi:10.48550/arxiv.2408.00714 2024
[33]

Riba, E., Mishkin, D., Ponsa, D., Rublee, E., and Bradski, G. (2020). Kornia: an open source differentiable computer vision library for PyTorch. In 2020 IEEE Winter Conference on Applications of Computer Vision (WACV). pp. 3663–3672. doi:10.1109/WACV45572. 2020.9093363

work page doi:10.1109/wacv45572 2020
[34]

Bradski, G. (2000). The OpenCV library. Dr. Dobb’s Journal of Software Tools

2000
[35]

Gillies, S., van der Wel, C., Van den Bossche, J., Taves, M.W., Arnott, J., Ward, B.C. et al. (2007). Shapely. GitHub. URL:https://github.com/shapely/shapely. doi: 10.5281/zenodo.5597138

work page doi:10.5281/zenodo.5597138 2007
[36]

Gillies, S. et al. (2013). Rasterio: geospatial raster i/o for Python programmers. GitHub. URL:https://github.com/rasterio/rasterio

2013
[37]

Zwietering, M.H., Jongenburger, I., Rombouts, F .M., and van ’t Riet, K. (1990). Modeling of the bacterial growth curve. Applied and Environmental Microbiology56, 1875–1881. doi: 10.1128/aem.56.6.1875-1881.1990

work page doi:10.1128/aem.56.6.1875-1881.1990 1990
[38]

SciPy1.0: FundamentalalgorithmsforscientificcomputinginPython,

Virtanen, P ., Gommers, R., Oliphant, T.E., Haberland, M., Reddy, T., Cournapeau, D., Burovski, E., Peterson, P ., Weckesser, W., Bright, J., van der Walt, S.J., Brett, M., Wilson, J., Millman, K.J., Mayorov, N., Nelson, A.R.J., Jones, E., Kern, R., Larson, E., Carey, C.J., Po- lat, ˙I., Feng, Y ., Moore, E.W., VanderPlas, J., Laxalde, D., Perktold, J., C...

work page doi:10.1038/s41592-019-0686-2 2020

[1] [1]

Wang, P ., Robert, L., Pelletier, J., Dang, W.L., Taddei, F ., Wright, A., and Jun, S. (2010). Robust Growth of Escherichia coli. Current Biology20, 1099–1103. doi:10.1016/j.cub. 2010.04.045

work page doi:10.1016/j.cub 2010

[2] [2]

Gr ¨unberger, A., Paczia, N., Probst, C., Schendzielorz, G., Eggeling, L., Noack, S., Wiechert, W., and Kohlheyer, D. (2012). A disposable picolitre bioreactor for cultivation and investiga- tion of industrially relevant bacteria on the single cell level. Lab on a Chip12, 2060. doi: 10.1039/c2lc40156h

work page doi:10.1039/c2lc40156h 2012

[3] [5]

Kaiser, M., Jug, F ., Julou, T., Deshpande, S., Pfohl, T., Silander, O.K., Myers, G., and van Nimwegen, E. (2018). Monitoring single-cell gene regulation under dynamically controllable conditions with integrated microfluidics and software. Nature Communications9, 212. doi: 10.1038/s41467-017-02505-0

work page doi:10.1038/s41467-017-02505-0 2018

[4] [6]

Gr ¨unberger, A., Probst, C., Helfrich, S., Nanda, A., Stute, B., Wiechert, W., von Lieres, E., N¨oh, K., Frunzke, J., and Kohlheyer, D. (2015). Spatiotemporal microbial single-cell analysis using a high-throughput microfluidics cultivation platform. Cytometry Part A87, 1101–1115. doi:10.1002/cyto.a.22779

work page doi:10.1002/cyto.a.22779 2015

[5] [7]

Helfrich, S., Pfeifer, E., Kr ¨amer, C., Sachs, C.C., Wiechert, W., Kohlheyer, D., N¨oh, K., and Frunzke, J. (2015). Live cell imaging of SOS and prophage dynamics in isogenic bacterial populations. Molecular Microbiology98, 636–650. doi:10.1111/mmi.13147

work page doi:10.1111/mmi.13147 2015

[6] [8]

Sachs, C.C., Gr ¨unberger, A., Helfrich, S., Probst, C., Wiechert, W., Kohlheyer, D., and N¨oh, K. (2016). Image-based single cell profiling: High-throughput processing of mother machine experiments. PLOS ONE11, e0163453. doi:10.1371/journal.pone.0163453

work page doi:10.1371/journal.pone.0163453 2016

[7] [9]

Bl ¨obaum, L., T¨auber, S., and Gr¨unberger, A. (2023). Protocol to perform dynamic microflu- idic single-cell cultivation of C. glutamicum. STAR Protocols4, 102436. doi:10.1016/j. xpro.2023.102436

work page doi:10.1016/j 2023

[8] [10]

O’Connor, O.M., Alnahhas, R.N., Lugagne, J.B., and Dunlop, M.J. (2022). DeLTA 2.0: A deep learning pipeline for quantifying single-cell spatial and temporal dynamics. PLOS Computational Biology18, e1009797. doi:10.1371/journal.pcbi.1009797. 19

work page doi:10.1371/journal.pcbi.1009797 2022

[9] [11]

Prangemeier, T., Wildner, C., Franc ¸ani, A.O., Reich, C., and Koeppl, H. (2022). Y east cell segmentation in microstructured environments with deep learning. Biosystems211, 104557. doi:10.1016/j.biosystems.2021.104557

work page doi:10.1016/j.biosystems.2021.104557 2022

[10] [12]

Zhou, S., Chen, B., Fu, E.S., and Y an, H. (2023). Computer vision meets microfluidics: a label-free method for high-throughput cell analysis. Microsystems & Nanoengineering9. doi:10.1038/s41378-023-00562-8

work page doi:10.1038/s41378-023-00562-8 2023

[11] [13]

Cutler, K.J., Stringer, C., Lo, T.W., Rappez, L., Stroustrup, N., Brook Peterson, S., Wiggins, P .A., and Mougous, J.D. (2022). Omnipose: a high-precision morphology- independent solution for bacterial cell segmentation. Nature Methods pp. 1–11. doi: 10.1038/s41592-022-01639-4

work page doi:10.1038/s41592-022-01639-4 2022

[12] [14]

Stringer, C., Wang, T., Michaelos, M., and Pachitariu, M. (2020). Cellpose: a gener- alist algorithm for cellular segmentation. Nature Methods18, 100–106. doi:10.1038/ s41592-020-01018-x

2020

[13] [15]

Pachitariu, M., Rariden, M., and Stringer, C. (2025). Cellpose-SAM: superhuman general- ization for cellular segmentation. bioRxiv. doi:10.1101/2025.04.28.651001

work page doi:10.1101/2025.04.28.651001 2025

[14] [16]

Marks, U., Israel, R., Van Valen, D. et al. (2025). A foundation model for cellular segmenta- tion. Nature Methods. doi:10.1038/s41592-025-02879-w

work page doi:10.1038/s41592-025-02879-w 2025

[15] [17]

Archit, A., Freckmann, L., Nair, S., Khalid, N., Hilt, P ., Rajashekar, V., Freitag, M., Teuber, C., Buckley, G., von Haaren, S., Gupta, S., Dengel, A., Ahmed, S., and Pape, C. (2025). Segment anything for microscopy. Nature Methods. doi:10.1038/s41592-024-02580-4

work page doi:10.1038/s41592-024-02580-4 2025

[16] [18]

Kasahara, K., Seiffarth, J., Stute, B., von Lieres, E., Drepper, T., N¨oh, K., and Kohlheyer, D. (2025). Unveiling microbial single-cell growth dynamics under rapid periodic oxygen oscil- lations. Lab on a Chip25, 2234–2246. doi:10.1039/D5LC00065C

work page doi:10.1039/d5lc00065c 2025

[17] [19]

Witting, L., Seiffarth, J., Stute, B., Schulze, T., Hofer, J.M., N ¨oh, K., Eisenhut, M., Weber, A.P .M., von Lieres, E., and Kohlheyer, D. (2025). A microfluidic system for the cultivation of cyanobacteria with precise light intensity and CO2 control: enabling growth data acquisition at single-cell resolution. Lab on a Chip25, 319–329. doi:10.1039/D4LC00567H

work page doi:10.1039/d4lc00567h 2025

[18] [20]

Dal Co, A., van Vliet, S., Kiviet, D.J., Schlegel, S., and Ackermann, M. (2020). Short-range interactions govern the dynamics and functions of microbial communities. Nature Ecology & Evolution4, 366–375. doi:10.1038/s41559-019-1080-2

work page doi:10.1038/s41559-019-1080-2 2020

[19] [21]

Thiermann, R., Sandler, M., Ahir, G., Sauls, J.T., Schroeder, J., Brown, S., Le Treut, G., Si, F ., Li, D., Wang, J.D., and Jun, S. (2024). Tools and methods for high-throughput single-cell imaging with the mother machine. eLife12, RP88463. doi:10.7554/eLife.88463. 20

work page doi:10.7554/elife.88463 2024

[20] [22]

Lugagne, J.B., Blassick, C.M., and Dunlop, M.J. (2024). Deep model predictive control of gene expression in thousands of single cells. Nature Communications15. doi:10.1038/ s41467-024-46361-1

2024

[21] [23]

Merrin, J. (2019). Frontiers in Microfluidics, a Teaching Resource Review. Bioengineering 6, 109. doi:10.3390/bioengineering6040109

work page doi:10.3390/bioengineering6040109 2019

[22] [24]

Long, Z., Nugent, E., Javer, A., Cicuta, P ., Sclavi, B., Lagomarsino, M.C., and Dorfman, K.D. (2013). Microfluidic chemostat for measuring single cell dynamics in bacteria. Lab on a Chip13, 947–954. doi:10.1039/C2LC41196B

work page doi:10.1039/c2lc41196b 2013

[23] [25]

Shi, Y ., Tabet, J.S., Milkie, D.E., Daugird, T.A., Y ang, C.Q., Ritter, A.T., Giovannucci, A., and Legant, W.R. (2024). Smart lattice light-sheet microscopy for imaging rare and complex cellular events. Nature Methods. doi:10.1038/s41592-023-02126-0

work page doi:10.1038/s41592-023-02126-0 2024

[24] [26]

Waithe, D., Brown, J.M., Reglinski, K., Diez-Sevilla, I., Roberts, D., and Eggeling, C. (2020). Object detection networks and augmented reality for cellular detection in fluorescence mi- croscopy. Journal of Cell Biology219. doi:10.1083/jcb.201903166

work page doi:10.1083/jcb.201903166 2020

[25] [27]

Al-Hamadani, M.N.A., Poroszlay, R., Szeman-Nagy, G., Hajdu, A., Hadjidemetriou, S., Ferrarini, L., and Harangi, B. (2025). Improving cell detection and tracking in microscopy images using YOLO and an enhanced DeepSORT algorithm. Sensors25, 4361. doi: 10.3390/s25144361

work page doi:10.3390/s25144361 2025

[26] [28]

Jocher, G., Qiu, J., and Chaurasia, A. (2023). Ultralytics YOLO. GitHub. URL:https: //github.com/ultralytics/ultralytics. doi:10.5281/zenodo.7347926

work page doi:10.5281/zenodo.7347926 2023

[27] [29]

Chiron, L., Le Bec, M., Cordier, C., Pouzet, S., Milunov, D., Banderas, A., Di Meglio, J.M., Sorre, B., and Hersen, P . (2022). CyberSco.Py an open-source software for event-based, conditional microscopy. Scientific Reports12, 11579. doi:10.1038/s41598-022-15207-5

work page doi:10.1038/s41598-022-15207-5 2022

[28] [30]

Passmore, J.B., Rates, A., Schr ¨oder, J., van Laarhoven, M.T.P ., Hellebrekers, V.J.W., van Hoef, H.G., Geurts, A.J.M., van Straaten, W., Nijenhuis, W., Berger, F ., Smith, C.S., Smal, I., and Kapitein, L.C. (2025). Closed-loop optogenetic control of cell biology enables outcome- driven microscopy. Nature Communications17, 1087. doi:10.1038/s41467-025-67848-5

work page doi:10.1038/s41467-025-67848-5 2025

[29] [31]

Gruenberger, A., Probst, C., Heyer, A., Wiechert, W., Frunzke, J., and Kohlheyer, D. (2013). Microfluidic picoliter bioreactor for microbial single-cell analysis: Fabrication, system setup, and operation. Journal of Visualized Experiments. URL:https://app.jove.com/t/50560. doi:10.3791/50560

work page doi:10.3791/50560 2013

[30] [32]

Ravi, N., Gabeur, V., Hu, Y .T., Hu, R., Ryali, C., Ma, T., Khedr, H., R ¨adle, R., Rolland, C., Gustafson, L., Mintun, E., Pan, J., Alwala, K.V., Carion, N., Wu, C.Y ., Girshick, R., Doll´ar, P ., and Feichtenhofer, C. (2024). SAM 2: Segment Anything in Images and Videos. arXiv. doi: 10.48550/arXiv.2408.00714. 21

work page internal anchor Pith review Pith/arXiv arXiv doi:10.48550/arxiv.2408.00714 2024

[31] [33]

Riba, E., Mishkin, D., Ponsa, D., Rublee, E., and Bradski, G. (2020). Kornia: an open source differentiable computer vision library for PyTorch. In 2020 IEEE Winter Conference on Applications of Computer Vision (WACV). pp. 3663–3672. doi:10.1109/WACV45572. 2020.9093363

work page doi:10.1109/wacv45572 2020

[32] [34]

Bradski, G. (2000). The OpenCV library. Dr. Dobb’s Journal of Software Tools

2000

[33] [35]

Gillies, S., van der Wel, C., Van den Bossche, J., Taves, M.W., Arnott, J., Ward, B.C. et al. (2007). Shapely. GitHub. URL:https://github.com/shapely/shapely. doi: 10.5281/zenodo.5597138

work page doi:10.5281/zenodo.5597138 2007

[34] [36]

Gillies, S. et al. (2013). Rasterio: geospatial raster i/o for Python programmers. GitHub. URL:https://github.com/rasterio/rasterio

2013

[35] [37]

Zwietering, M.H., Jongenburger, I., Rombouts, F .M., and van ’t Riet, K. (1990). Modeling of the bacterial growth curve. Applied and Environmental Microbiology56, 1875–1881. doi: 10.1128/aem.56.6.1875-1881.1990

work page doi:10.1128/aem.56.6.1875-1881.1990 1990

[36] [38]

SciPy1.0: FundamentalalgorithmsforscientificcomputinginPython,

Virtanen, P ., Gommers, R., Oliphant, T.E., Haberland, M., Reddy, T., Cournapeau, D., Burovski, E., Peterson, P ., Weckesser, W., Bright, J., van der Walt, S.J., Brett, M., Wilson, J., Millman, K.J., Mayorov, N., Nelson, A.R.J., Jones, E., Kern, R., Larson, E., Carey, C.J., Po- lat, ˙I., Feng, Y ., Moore, E.W., VanderPlas, J., Laxalde, D., Perktold, J., C...

work page doi:10.1038/s41592-019-0686-2 2020