Shape Factor Analysis as a Quantitative Framework for Assessing Spheroid and Organoid Morphology and Invasiveness

This paper introduces a custom MATLAB-based shape factor analysis framework, specifically utilizing radial length variance, to provide a more reliable and comprehensive quantitative method for assessing spheroid and organoid morphology and invasiveness compared to standard shape descriptors.

The Gist

Imagine you are a detective trying to solve a mystery, but instead of looking for fingerprints, you are looking at tiny, 3D balls of cells growing in a petri dish. These balls, called spheroids and organoids, are like miniature models of human organs (like a tiny breast or a tiny intestine) that scientists use to study diseases like cancer.

The big problem? It's hard to tell the difference between a healthy, calm ball of cells and a sick, aggressive ball of cells just by looking at them. A healthy ball is usually round and smooth, like a marble. A cancerous ball often starts to sprout jagged fingers or spikes as it tries to invade its surroundings, looking more like a starfish or a sea urchin.

Until now, scientists had to guess or use very complicated, expensive computer programs to measure these shapes. This paper introduces a new, simpler, and smarter way to measure these "cell balls" using a concept called Shape Factor Analysis.

Here is how the paper breaks it down, using some everyday analogies:

1. The "Rubber Band" Test (Standard Shape Descriptors)

Think of the standard way scientists measure these balls as putting a rubber band around the shape.

• Circularity: How close is the shape to a perfect circle? If the rubber band is tight and round, it's healthy. If the rubber band has to stretch over jagged spikes, the "circularity" score drops.
• Solidity: Imagine filling the shape with water. If the shape is a solid block, it holds a lot of water. If it has deep cracks or holes (like a crescent moon), it holds less water relative to its size.

The Catch: These standard tools are a bit like a blunt instrument. They can tell you a shape isn't a perfect circle, but they can't always tell why. Is it because the ball is just squashed (elongated)? Or is it because it has dangerous, invasive spikes? Sometimes, a squashed ball looks just as "bad" as an invasive one, leading to confusion.

2. The "Spaghetti String" Test (Radial Length Analysis)

This is the paper's big innovation. Instead of just looking at the whole shape, the authors created a custom computer program (using MATLAB) that acts like a spider spinning a web.

Imagine the center of the cell ball is a spider. The program shoots out hundreds of tiny strings (radial lines) from the center to every single point on the edge of the ball.

• The Measurement: It measures the length of every single string.
• The Math:
  • If the ball is a perfect circle, all the strings are the exact same length.
  • If the ball has spikes (invasion), some strings are super long (hitting the tips of the spikes) and some are short (hitting the valleys between spikes).
  • The program calculates the Standard Deviation (how much the lengths vary) and counts the Crossings (how many times the strings jump from being "too long" to "too short").

Why this is better: This method is like a metal detector. It doesn't just say "this shape is weird"; it specifically counts the number of "spikes" or "fingers" the cancer cells are growing. It can tell the difference between a ball that is just squashed (all strings are roughly the same length, just longer in one direction) and a ball that is actively attacking its neighbors (strings vary wildly in length).

3. Real-World Applications

The authors tested this "Spaghetti String" method on three different scenarios:

• Digital Phantoms (The Practice Run): They drew perfect shapes on a computer (circles, stars, squares) to prove their math worked. The standard tools got confused by squares and triangles, but the new method correctly identified the "spiky" ones.
• Organoids (The Tiny Organs): They looked at tiny models of uterine and fallopian tube cells. They found that cancerous organoids had "folds" and "bumps" inside them. The new method could spot these subtle folds better than the old tools, acting like a high-resolution map of the terrain.
• Drug Testing (The Cure): They treated cancer cells with drugs to stop them from invading. The old method (just measuring the total area) took a long time to notice the drugs were working. The new method noticed the spikes disappearing almost immediately. It's like noticing a weed has stopped growing its thorns before you even see the whole plant shrink.

The Bottom Line

This paper is essentially giving scientists a new pair of glasses.

• Old Glasses: "That cell ball looks a bit weird and not very round."
• New Glasses (Shape Factor Analysis): "That cell ball has 15 distinct invasive spikes, and they are growing 20% faster than yesterday. Also, this drug is successfully cutting those spikes off."

By turning the visual shape of cells into simple numbers, this method allows researchers to screen thousands of drugs faster, catch cancer earlier, and understand how diseases spread, all without needing expensive, super-complex AI or staring at microscopes for hours. It brings the precision of a clinical radiologist (who looks at X-rays to find tumors) right into the petri dish.

Technical Summary

1. Problem Statement

Advanced 3D cell culture models, such as spheroids and organoids, are critical for studying tissue microenvironments, disease progression (particularly cancer invasion), and drug screening. However, a significant bottleneck exists in quantitatively assessing their morphology.

• Current Limitations: Researchers often rely on qualitative visual inspection or simple metrics like total area or diameter. These fail to distinguish between a growing non-invasive spheroid and an invasive one.
• Existing Tools: Standard image analysis software (e.g., ImageJ/FIJI) offers shape descriptors (circularity, solidity, aspect ratio), but these have limitations:
  • They often cannot differentiate between elongation, concavity, and specific invasive protrusions (fingers/tips).
  • Machine learning approaches exist but require specialized software and expertise, limiting high-throughput adoption.
• Clinical Gap: In clinical radiology, tumor margin irregularity is a key prognostic marker for malignancy, yet this quantitative rigor is largely missing in in vitro research.

2. Methodology

The authors developed a quantitative framework combining standard shape descriptors with a custom Radial Length Analysis algorithm to objectively measure morphological changes.

• Digital Phantoms: Ten custom digital shapes were created to test the sensitivity of various metrics against specific morphological features (smooth contours, protrusions, concavities, and budding).
• Standard Shape Descriptors (FIJI/ImageJ): The study evaluated dimensionless metrics including:
  • Circularity: $4\pi \times (\text{Area} / \text{Perimeter}^2)$
  • Solidity: $\text{Area} / \text{Convex Area}$
  • Convexity: $\text{Convex Perimeter} / \text{Perimeter}$
  • Roundness/Aspect Ratio: Measures of elongation.
• Radial Length Analysis (Custom MATLAB/FIJI Pipeline):
  • Inspired by clinical breast imaging (BI-RADS system), this method calculates the Euclidean distance from the shape's centroid to every point on its perimeter.
  • Key Metrics:
    1. Standard Deviation of Radial Lengths (SDRL): Quantifies the magnitude of perimeter variations (size-dependent).
    2. Average Radial Length Crossings (ARLC): Counts how many times the radial length vector crosses the average radial length. This quantifies the frequency of undulations or protrusions (size-independent).
  • Tunability: The algorithm includes parameters for "excursion length" (consecutive points above/below average) and "minimum crossing distance" to filter noise and adapt to specific image resolutions.
  • Segmented Analysis: The code allows for analyzing specific angular segments ("pie slices") to detect asymmetric invasion.
• Experimental Validation: The framework was applied to:
  • Breast Cancer Spheroids: Embedded in collagen/Matrigel, imaged via bright-field and confocal microscopy.
  • Organoids: Endometrial and oviductal epithelial organoids (both normal and cancer-induced).
  • Drug Screening: Spheroids treated with matrix metalloproteinase and contractility inhibitors.

3. Key Contributions

• Development of a Custom Pipeline: Created a user-friendly, open-source workflow (FIJI macros + MATLAB scripts) to calculate ARLC and SDRL, bridging the gap between clinical radiology techniques and in vitro research.
• Comprehensive Metric Comparison: Systematically demonstrated that no single standard shape descriptor is universally sufficient.
  • Circularity is best for detecting isolated invasion tips.
  • Solidity/Convexity are better for detecting folds or concave regions (common in organoids).
  • Roundness only detects elongation, not invasion.
• Introduction of ARLC: Established ARLC as a robust, size-independent metric specifically for quantifying the number of collective invasion protrusions, which standard descriptors often miss or conflate with elongation.
• Asymmetry Detection: Demonstrated the ability to segment spheroids into angular regions to quantify directional (asymmetric) invasion driven by microenvironmental gradients.

4. Key Results

• Digital Phantom Analysis:
  • Standard descriptors (Solidity, Roundness) failed to distinguish between shapes with different numbers of protrusions if the overall area remained similar.
  • Circularity and Convexity were most sensitive to protrusions.
  • ARLC successfully increased with the number of protrusions, while SDRL increased with the length of protrusions.
• Spheroid Invasion:
  • Invasive spheroids showed significant decreases in circularity, solidity, and convexity compared to non-invasive controls.
  • ARLC and SDRL increased significantly as early as Day 1 of culture, detecting invasive phenotypes earlier and more specifically than simple area measurements.
  • Radial analysis successfully distinguished between elongated non-invasive spheroids (high SDRL, low ARLC) and round invasive spheroids (high SDRL, high ARLC).
• Organoid Morphology:
  • In endometrial organoids, Circularity was the most effective metric for distinguishing uniform shapes from irregular, folded, or lobulated shapes.
  • In oviductal organoids, cancer-induced lumens with internal cell clusters (concave structures) showed significantly lower Convexity and Solidity compared to controls.
• Drug Screening Application:
  • Treatment with inhibitors (Batimastat + Y27632) significantly reduced ARLC and SDRL at Day 1, indicating the suppression of collective invasion tips.
  • The dimensionless nature of ARLC allowed for direct comparison of data across different magnifications (e.g., comparing Day 1 high-mag images to Day 2 low-mag images), a critical feature for high-throughput screening where spheroid sizes vary drastically.

5. Significance and Impact

• High-Throughput Compatibility: The method works with low-resolution bright-field images, enabling longitudinal monitoring of live spheroids without the cost and time of high-resolution confocal imaging or staining.
• Standardization: Provides a standardized, quantitative language for describing spheroid/organoid morphology, reducing reliance on subjective visual scoring.
• Translational Bridge: By adapting clinical radiology metrics (margin irregularity) to in vitro models, the study creates a direct link between bench research and clinical prognostic indicators.
• Drug Discovery: Enables earlier and more sensitive detection of drug efficacy in suppressing invasion, potentially accelerating pre-clinical screening workflows.
• Open Science: All code (FIJI macros and MATLAB scripts) is publicly available on GitHub, facilitating immediate adoption and adaptation by the broader research community.

In conclusion, the paper argues that a multivariate approach—combining standard dimensionless shape factors with custom radial length analysis—is essential for the accurate, reliable, and high-throughput phenotypic profiling of 3D tissue models.