Subono, Sholeh Hadi Pramono, Erni Yudaningtyas, Muhammad Aziz Muslim
This study presents a Deep Convolutional Generative Adversarial Network (DCGAN) with multiclass classifiers for fetal heart rate (FHR) classification, focusing on imbalanced data between normal, suspected, and pathological cases. Using the CTU-UHB dataset, this study yielded model performance in terms of precision, recall, F1 score, and overall accuracy. The DCGAN model successfully improved the class balance, particularly for the underrepresented pathological cases. The model classification achieved high accuracy (0.97), with strong performance in detecting 'Normal' and 'Fetal Distress' cases, although the precision for the 'suspected' class was slightly lower (0.90). The other graphs also show significant results, especially the detection accuracy and loss, the higher the epoch, the closer the accuracy is to 1.00 and the loss is to 0.00. The confusion matrix results obtained 328 'normal' data, 60 'suspect' data and 27 'pathology' data. These results demonstrate the effectiveness of DCGAN in reducing data imbalance while maintaining high classification accuracy, thus showing potential for improving foetal hypoxia detection in medical applications. © 2025 IEEE.
Universitas Brawijaya, Malang, Indonesia; Politeknik Negeri Banyuwangi, Business and Informatics Department, Banyuwangi, Indonesia
