Paper Title
A Comparative Evaluation of Classical Machine Learning Algorithms for Sperm Morphology Classification on the MHSMA Dataset

Abstract
Infertility represents a growing global health concern, with male factors accounting for nearly half of all cases. Among the diagnostic parameters, sperm morphology plays a critical role in evaluating sperm quality and reproductive potential. Traditional manual assessments of sperm morphology, though widely practiced, are time-intensive, subject to human error, and prone to inter-observer variability. To address these limitations, this study investigates the application of Machine Learning (ML) techniques for automated classification of sperm morphological abnormalities using the MHSMA (Modified Human Sperm Morphology Analysis) dataset. The dataset comprises grayscale images of individual sperm cells, annotated by clinical experts to identify abnormalities across four primary morphological regions: acrosome, head, tail, and vacuole. Each image undergoes preprocessing and is subjected to binary classification across multiple ML models to determine the optimal algorithm for each morphological feature. The approach offers a scalable, objective alternative to manual evaluation. Additionally, the findings highlight the potential of AI-driven diagnostic systems to enhance accuracy, reduce workload, and improve the overall efficiency of male infertility diagnostics. Keywords - Sperm Morphology, Male Infertility, MHSMA Dataset, Machine Learning, Classification, Random Forest, SVM, KNN.