A statistically rigorous multi-scale texture analysis framework for 3D spheroid characterization: temporal...

Area of Science:

• Biomedical Engineering
• Computational Biology
• Cancer Research

Background:

• 3D tumor spheroids are valuable models for studying cancer but face imaging depth limitations.
• Current methods capture either limited single-cell resolution or discrete molecular snapshots, missing dynamic biological processes.

Purpose of the Study:

• To develop a computational framework for statistically rigorous, ensemble-level morphological profiling of 3D tumor spheroids.
• To enable continuous monitoring of spheroid dynamics and complement molecular endpoint assays.

Main Methods:

• Integrated multi-scale texture analysis (37 features) and global standardization.
• Employed autocorrelation-informed block bootstrap resampling for temporal dependence.
• Validated the framework using lung cancer spheroids and external RNA-sequencing data.

Main Results:

• Global standardization effectively normalized features while preserving biological signal (16.8-fold discrimination).
• Temporal autocorrelation analysis informed resampling for valid statistical inference.
• Morphological profiling showed quantitative concordance (within 15%) with molecular differences (VIM/CDH1 ratio).

Conclusions:

• The framework provides statistically valid, biologically grounded morphological profiling for 3D tumor spheroids.
• Enables continuous monitoring for applications like drug screening, organoid tracking, and biomanufacturing quality control.