One step beyond CASA Systems: Artificial intelligence (AI) and deep learning (DL) for canine sperm analysis (CSA)

Computer Assisted Sperm Analysis (CASA) systems have become indispensable tools for determinining objective motility (percentage) and precise kinematic parameters (velocities μm/sec) in mammalian semen analysis over the past decades. However, conventional CASA platforms dependent on grayscale image acquisition and predefined mathematical algorithms. These rigid frameworks are often inadequate for capturing the complex, dynamic movement patterns of spermatozoa. In canine reproduction, where semen quality asssessment critical for Breeding Soundness Examinations (BSEs), Assisted Reproductive Technologies (ART), and clinical decision making, these limitations may compromise the accuracy and clinical relevance of motility based diagnostics. Recent advancements in Artificial Intelligence (AI) and Deep Learning (DL) offer transformative opportunities to overcome the intrinsic constraints of conventional CASA systems. AI/DL based approaches enable data driven feature extraction, improved sperm detection and superior tracking capabilities. Unlike traditional CASAs, AI/DL integrated systems, having been recently developed provide robust classification of motility patterns without strict dependence on manually defined thresholds, reducing sensitivity to variables such as illumination, magnification and background extender and seminal plasma (SP) or spermatogenesis based particiles. Moreover, DL models can simultaneously integrate spatial, temporal and morphological information as specialized datasets expand. This article summarized the current state of the art in conventional (gray scale) and modern (AI/DL) canine CASA applications and critically discusses their methodological limitations based on studies published between 2013 and 2026, also covering researches and development pathway related with three successive projects: ESAS (2014-2018)_conventional/ESAS TİP (2018-2022)_AI-DL and ESAS TÜG_AI-DL (initiated 2022 and in progress) partly financially and scientifically supported by The Scientific and Technological Research Council of Türkiye (TÜBİTAK) and Technology and Innovation Funding Programmes Directorate (TEYDEB) (2013-2018 conventional, 2018-2026 AI/DL model R&D studies, especially 1507 and TÜBİTAK TEYDEB 1501 R&D Program (ESAS Commercial Product Development (TÜG) Project No: 3230063)). This current article highlights the technological transition from fragmented, parameter based motility assessments toward holistic, pattern oriented sperm evaluations, evaluating its ultimate implications for fertility prediction and routine canine veterinary andrology with one step beyond of conventional CASA with deep learning and artificial intelligence.