Machine learning driven semi-automated framework for yeast sporulation efficiency quantification using ilastik segmentation and Fiji nuclear enumeration

Accurate quantification of yeast sporulation efficiency is essential for genetic studies, but manual counting remains time-consuming and susceptible to subjective bias. Although deep learning tools like cellpose provide automated solutions, there exists a compelling need for alternative approaches that enable the quantification of spores. Our methodology employs ilastik's texture-feature optimization to reliably segment sporulating mother cells, intentionally avoiding explicit tetrad discrimination to ensure robustness across diverse spore morphologies. Subsequent Fiji-based image processing employs optimized algorithms for accurate spore quantification within cellular boundaries, facilitating automated batch classification of dyads, triads, and tetrads. Quantitative validation demonstrates our pipeline maintains strong concordance with manual counting (93.4 % agreement, ICC = 0.94) alongside a 68 % reduction in processing time (P < 0.001). The pipeline's reliability was further verified in Hsp82 phosphorylation mutants, consistently enables quantification of sporulation efficiency across genetic backgrounds. To balance throughput and precision, our workflow intentionally combines automated image processing (ilastik segmentation, Fiji quantification) with manual quality control checkpoints (segmentation validation). This modular pipeline allows adjustable segmentation parameters, compatibility with alternative nuclear markers, and batch processing of diverse imaging datasets. By combining accessibility with precision, our method provides laboratories a reproducible alternative to fully manual counting while maintaining compatibility with standard microscopy setups.