A Convolutional Neural Network Model with Feature Fusion for Sperm Morphology Classification
Abstract
Sperm morphology analysis is a key parameter in male fertility evaluation, but the manual process is subjective and time-consuming. Therefore, an automated deep learning classification system offers a potential solution. This study evaluates and compares MobileNetV2 and EfficientNetB0, individually and in combination via feature fusion, for sperm morphology image classification using the SMID dataset of 3000 images. MobileNetV2 and EfficientNetB0 models achieved accuracies of 62.75% and 33.06%, respectively, after 50 epochs, while the feature fusion model reached 85.04% in only 5 epochs. Thus, combining features from both architectures yields superior accuracy and efficiency for automated sperm analysis.
Fields of Study & Subjects
Sperm and Testicular Function
Reproductive Biology and Fertility
Ovarian function and disorders
Publication details
- Journal / source
- Engineering Technology & Applied Science Research
- Publisher
- Engineering, Technology & Applied Science Research
- Volume
- 16
- Issue
- 1
- Pages
- 31408–31413
- Publication date
- 9 February 2026