Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Enhancing Cell Counting through MLOps: A Structured Approach for Automated Cell Analysis

Matteo Testi, Luca Clissa, Matteo Ballabio, Salvatore Ricciardi, Federico Baldo, Emanuele Frontoni, Sara Moccia, Gennario Vessio

TL;DR

The paper tackles the challenge of reliable, scalable cell counting in fluorescence microscopy by introducing CC-MLOps, a structured MLOps framework that spans data access, preprocessing, model training, monitoring, explainability, and sustainability. It demonstrates a practical pre-production use-case using the yellow FNC v2 dataset and the cell-ResUNet architecture, comparing Dice and Focal losses and validating with CI/CD on GitHub/Vertex AI and monitoring via Neptune.ai. Key findings show improved reliability, reproducibility, and trust through continuous validation, explainability with Grad-CAM, and sustainability considerations via CodeCarbon, highlighting trade-offs between model performance and environmental impact. The work provides a concrete blueprint for researchers to deploy scalable, auditable, and regulatorily aware cell counting pipelines in neuroscience, pharmacology, and environmental monitoring, while acknowledging limitations and outlining paths toward production deployment and broader domain adaptation.

Abstract

Machine Learning (ML) models offer significant potential for advancing cell counting applications in neuroscience, medical research, pharmaceutical development, and environmental monitoring. However, implementing these models effectively requires robust operational frameworks. This paper introduces Cell Counting Machine Learning Operations (CC-MLOps), a comprehensive framework that streamlines the integration of ML in cell counting workflows. CC-MLOps encompasses data access and preprocessing, model training, monitoring, explainability features, and sustainability considerations. Through a practical use case, we demonstrate how MLOps principles can enhance model reliability, reduce human error, and enable scalable Cell Counting solutions. This work provides actionable guidance for researchers and laboratory professionals seeking to implement machine learning (ML)- powered cell counting systems.

Enhancing Cell Counting through MLOps: A Structured Approach for Automated Cell Analysis

TL;DR

The paper tackles the challenge of reliable, scalable cell counting in fluorescence microscopy by introducing CC-MLOps, a structured MLOps framework that spans data access, preprocessing, model training, monitoring, explainability, and sustainability. It demonstrates a practical pre-production use-case using the yellow FNC v2 dataset and the cell-ResUNet architecture, comparing Dice and Focal losses and validating with CI/CD on GitHub/Vertex AI and monitoring via Neptune.ai. Key findings show improved reliability, reproducibility, and trust through continuous validation, explainability with Grad-CAM, and sustainability considerations via CodeCarbon, highlighting trade-offs between model performance and environmental impact. The work provides a concrete blueprint for researchers to deploy scalable, auditable, and regulatorily aware cell counting pipelines in neuroscience, pharmacology, and environmental monitoring, while acknowledging limitations and outlining paths toward production deployment and broader domain adaptation.

Abstract

Machine Learning (ML) models offer significant potential for advancing cell counting applications in neuroscience, medical research, pharmaceutical development, and environmental monitoring. However, implementing these models effectively requires robust operational frameworks. This paper introduces Cell Counting Machine Learning Operations (CC-MLOps), a comprehensive framework that streamlines the integration of ML in cell counting workflows. CC-MLOps encompasses data access and preprocessing, model training, monitoring, explainability features, and sustainability considerations. Through a practical use case, we demonstrate how MLOps principles can enhance model reliability, reduce human error, and enable scalable Cell Counting solutions. This work provides actionable guidance for researchers and laboratory professionals seeking to implement machine learning (ML)- powered cell counting systems.
Paper Structure (15 sections, 4 figures, 1 table)

This paper contains 15 sections, 4 figures, 1 table.

Table of Contents

  1. Introduction
  2. Related Work
  3. Our Contribution
  4. CC-MLOps
  5. Problem Understanding
  6. Data access and pre-processing
  7. ML Methodology
  8. ML Training & Testing
  9. Continuous Integration for MLOps in Cell Counting
  10. Continuous Delivery for MLOps in Cell Counting
  11. Continuous Training
  12. Continuous Monitoring
  13. Explainability
  14. Sustainability
  15. Conclusion

Figures (4)

  • Figure 1: The CC-MLOps life-cycle. This cycle begins with understanding the problem and concludes with sustainability, encompassing solution design, integration, and deployment. It represents a continuous journey of improving and refining ML models for reliable and accurate results.
  • Figure 2: FNV v2 - yellow sample images: these pictures depict brain tissues observed through fluorescence microscopy. Neuronal cells appear as yellow stains of different shape, size, and orientation, showing high variability in color features.
  • Figure 3: A visual overview of CI/CD pipelines for CC-MLOps leveraging Google Cloud and Vertex AI. It is illustrated data ingestion, automated model training, continuous delivery, and model monitoring for retraining
  • Figure 4: Grad-CAM heatmaps. From left to right, the figure depicts the original image followed by the Grad-CAM visualizations for the models trained with dice loss and focal loss, respectively.