Object Clas­si­fi­ca­tion in Mam­malian Cell Colony, Spher­oid, and Organoid Counting

In the study of cell aggregates—whether 2D or 3D cell colonies, spheroids, or organoids—researchers face a critical challenge: distinguishing biologically meaningful structures from artifacts. Automated imaging and analysis systems have revolutionized this process, offering high-throughput analysis while minimizing human bias. However, the accuracy of these systems hinges on carefully calibrated inclusion and exclusion criteria. This article explores how advanced platforms balance sensitivity and specificity to deliver reliable data, with a focus on applications across 2D and 3D cultures.

Automated colony counting systems are only as reliable as the parameters defining what constitutes a countable object. For colonies, spheroids, and organoids, these criteria must account for structural complexity, variability in size and morphology, and experimental context.

Biological relevance
Overcounting debris or undercounting legitimate small cell aggregates can distort conclusions. For example, in clonogenic assays, excluding colonies under 50 cells (<50 µm) ensures only viable clusters inform survival rate calculations. Conversely, in organoid drug screens, overly strict size thresholds might exclude physiologically relevant but irregularly shaped structures.

Reproducibility
Consistent thresholds are essential for cross-experiment comparisons. A study in PLOS ONE demonstrated that standardized detection boundaries reduced inter-lab variability in colony counts by 30%.

Impact on treatment response data
In radiation biology, lenient inclusion criteria may inflate survival rates by counting debris or undersized colonies or spheroids. These nuances underscore the need for platform flexibility, enabling researchers to tailor criteria to their specific models.

Modern cell aggregate imaging systems employ multi-parametric algorithms to evaluate objects, combining morphological, optical, and user-defined thresholds.

Key detection parameters include,

Size thresholds
Minimum diameter thresholds exclude debris (e.g., <100 µm for 3D colonies in soft agar), while maximum diameter thresholds flag oversized aggregates (e.g., >1,200 µm) that may represent fused spheroids or organoids.

Morphological filters
In using a circularity filter the detection of symmetrical structures (e.g., spheroids with circularity ≥0.7) can be prioritized over irregular debris. Or an edge sharpness filter can prioritize objects with well-defined boundaries that indicate viable aggregates, while suppressing objects with blurred edges that may signal aggregate disintegration.

Optical properties
Hardware capable of high-contrast imaging is capable of differentiating low-density organoids from background noise, while aggregate exclusion based on overall object optical density enables exclusion of partially degraded structures.

2D vs 3D analysis challenges
While 2D colonies are often straightforward (albeit time consuming) to analyze using standard counting methods, 3D cultures introduce depth-related complexities and generally a requirement to use a microscope, dramatically increasing the challenge and limiting throughput.

Spheroids and organoids grow across 3 dimensions, requiring systems with either sophisticated and processing-intensive z-stacking techniques or extended depth-of-field imaging to resolve overlapping structures in the z-axis. Most manual counting of 3D cell aggregates involves a highly trained individual painstakingly counting objects in one plane then moving to the next plane to count the next round of objects. This is a highly time-consuming and error-prone task.

2D colonies
Adherent clusters in adherent cultures benefit from size-based gating. For example, excluding objects <100 µm ensures only multicellular colonies are counted in CFU assays. In addition, having access to colony size data can point to treatment effects even if object counts are not affected. 

3D colonies and spheroids
Irregular growth patterns necessitate adaptive thresholds which can exclude or include misshapen cell aggregates. 

Organoids
Structural heterogeneity (e.g., intestinal vs. cerebral organoids) requires customizable parameters. Pixel density thresholds can exclude hollow or collapsed structures while retaining intact organoids.

Viability indicators
Automated systems often integrate into workflows for secondary parameters like phase-contrast texture analysis or fluorescence intensity (e.g., Calcein-AM staining) to exclude non-viable aggregates. This is critical in long-term organoid cultures where necrosis may increase over time.

Non-cellular material—such as fibrin strands in Matrigel-embedded organoids or cell-free clusters in suspension cultures—can skew counts. Advanced systems address this through:

Contrast-based exclusion
Low-intensity objects are omitted using adaptive thresholds.

Morphological profiling
Asymmetrical shapes (e.g., debris with high fractal dimensions) are auto excluded.

Handling overlaps and clusters
Cluster separation algorithms are vital for dense cultures. Advanced image processing algorithms achieve precision in resolving overlapping colonies, spheroids, and organoids. Even in high aggregating cell lines, preset algorithms will detect and size overlapping objects efficiently and across an entire study. 

While many automated platforms offer object classification, the GelCount system exemplifies how tailored workflows enhance reproducibility across 2D and 3D models.

GelCount™ employs high-depth-of-field imaging to view entire wells in a single pass, avoiding z-stacking complexity. As such it can clearly see all items in a well or plate at up to 5mm of depth. Its detection logic applies distinct criteria based on sample type.

In the case of adherent cell colonies growing in 2-dimensions, size thresholds (100–2,000 µm) and circularity filters prioritize adherent clusters.

For 3-dimensional cell aggregates such as spheroids or organoids, adaptive contrast adjustment compensates for depth-related light scattering, while morphology filters exclude irregular debris or non-specific objects and aggregates.

A key feature of GelCount™ is its ability to resolve overlapping structures through user-defined separation thresholds. Users can specify whether closely clustered aggregates should be classified as a single object or separate entities. For example, in a leukemia colony assay, setting a small gap threshold ensures overlapping colonies are counted individually. This logic remains consistent across all samples in a study, eliminating intra-experiment variability.

To further standardize counts, GelCount™ allows users to define a custom counting area (mask) within each well or plate. This excludes edge regions where aggregates often cluster unpredictably due to uneven settling. By analyzing the same central area across all replicates - for instance, 90% of the well - researchers mitigate bias caused by spatial heterogeneity. This is particularly critical for cell lines prone to over-aggregation, such as pancreatic cancer spheroids, where edge crowding complicates manual and automated counts alike.

Users can reprocess images offline, refining thresholds without re-imaging the original cultures. GelCount software is provided license-free and can be installed on any number of laboratory computers. Users can run the imager on one PC and process and analyse resulting images on a separate PC wherever they wish.  

Colony formation assays
GelCount™ detects microcolonies (>50 cells) post-irradiation, resolving overlaps. The System has been tested and has shown consistently <5% counting variability where manual counts can be variable upwards of 30-50%.

3D drug or therapy screening
Organoid, colony or spheroid studies utilize the GelCount™ to quantify the efficacy of drug regimes on, applying pixel density thresholds to exclude hollow structures. The system’s intrinsic generation of object diameter data can further facilitated dose-response modeling beyond just counts.

Library screen protocols
Library screens frequently comprise researchers looking at 10’s to 1000’s of compounds or drugs often in triplicates (or greater). The GelCount’s high throughput nature makes it a perfect fit to both increase plates processed and fully standardize these assays. 

Misapplied criteria compromise data at multiple levels:

Cross-lab variability
Labs using different exclusion criteria may see different results. The GelCount allows users with the system to input the same criteria across labs to ensure this is resolved in multi-site studies. 

Publication challenges
Journals increasingly require detailed method sheets outlining inclusion/exclusion logic, as highlighted in Cell Reports’ 2024 editorial guidelines.

Automated cell aggregate analysis demands a balance between standardization and flexibility. By leveraging systems that support customizable inclusion/exclusion criteria—capable of handling both 2D and 3D models—researchers can ensure their data reflects biological reality rather than algorithmic limitations or relying on manual miscounts or biases.

Key takeaways:

• Size, morphology, and contrast thresholds must align with model-specific biology.
• Post-processing flexibility and standardized masks enhance reproducibility.
• Transparent reporting of criteria (e.g., overlap resolution thresholds) is critical for cross-study validation.

For researchers exploring 3D culture or wanting to optimize clonogenic assays, platforms such as GelCount™ demonstrate how advanced imaging and adaptive logic elevate data quality. 

Contact us to learn how our solutions can elevate your research.