Segmentation and quantification of testicular histology images using machine learning bioimage analysis tools; Ilastik and Fiji software

Abstract

Histomorphological and histochemical techniques are widely used in infertility studies to assess testicular damage, determine the mechanisms involved, investigate potential interventions strategies, monitor treatment response and prognosis. Testis, a primary male reproductive organ is a compartmentalized organ made up of several seminiferous tubules and supporting tissue. Hence, focal damage is common, and accordingly, making accurate and insightful deductions require careful analysis of almost the entire testis section. However, manual analysis of testis histology sections to extract quantifiable data is hectic, time-consuming, liable to bias and undetected patchy damages and inter-personal variability. To circumvent these challenges, we present a step-by-step workflow using free, open-source interactive machine learning-based bioimage analysis tools; ilastik and Fiji. Ilastik uses a random forest classifier to compute generic pixel or object features for image segmentation. The segmented images exported from ilastik are subsequently quantified in FIJI to extract data for statistical analysis.

• •
  A step-by-step workflow using free, open-source interactive machine learning-based bioimage analysis tools; ilastik and Fiji.
• •
  A semiautomated, reproducible, time saving, unbiased, and broad scope method for analysis and heterogeneous tissue images.
• •
  Extraction of quantifiable data from images for statistical analysis to make comprehensive conclusions.