Gonadal Stage Identification and Oocyte Size Measurement Using DL Integrated Ultrasonography in Asian Sea Bass Lates calcarifer (Bloch, 1790)

This study investigated gonadal stages in Asian sea bass ( Lates calcarifer ) using portable ultrasonography (Sonosite Edge II, 6–13 MHz), with histological validation of female (primary oocyte, early vitellogenic, ripe) and male (ripe) stages. Distinct echogenicity patterns differentiated sexes, and ultrasound‐based gonad measurements showed strong correlations with actual dimensions (ovary length R 2 = 0.9942, ovary width R 2 = 0.9886, testis length R 2 = 0.9857, testis width R 2 = 0.9793). Deep learning models achieved 99.9% accuracy for sexual maturity prediction (DenseNet‐161) and 99.8% for sex classification (VGG16). CNN regression with L1 and L2 regularization was used to predict oocyte diameters. The MAE (mean absolute error) was 6.31 µm for minimum and 5.20 µm for maximum oocyte diameter, which is less than 1.82% of the actual oocyte diameter (256–544 µm). These results, along with averaging over 10‐fold and small confidence intervals, indicate the robustness of the classification and regression methods. This raises the possibility of automating the classification of the fish reproductive stage by combining deep learning with ultrasound technologies and further developing contactless underwater ultrasonographic assessment of gonadal maturation in commercially important fin fish species. It also reduces the dependency on experts to infer sexual maturation details from ultrasound scans. The developed machine learning model is deployed in real‐time processor, Nvidia Orin, through which real‐time testing is carried out and the results are validated.