Scholarlink

<p>This research implements a transfer learning framework for the detection of fetal cardiac anomalies using ultrasound scan images. The research framework uses four datasets, FetalEcho_V0504, FetalEcho_V0505, FetalEcho_V0506, and FetalEcho_V0507. The first one is the synthetic image dataset developed using GAN. The next three datasets are developed using the image transformation framework where three transformation techniques were developed. The transfer learning framework developed sixteen classifiers using four datasets and four deep learning models. These FCA detection classifiers are demonstrating average precision, recall, accuracy and F1-score of 0.94, 0.93, 0.93 and 0.92 respectively and the AUC of the best classifier identified is 0.99. The significance of individual performance of each classifier is identified using Friedman statistical test and based on the significance the evaluation of classifiers are performed to identify the best classifier. </p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Advanced Blueprint for Fetal Cardiac Anomaly Detection Using Transfer Learning and Synthetic FetalEcho Datasets with Image Transformations

  • 期刊论文

  • 发表日期：

    2026年6月1日

查看全文
  • M. O. Divya,
  • M. S. VIjaya

摘要

This research implements a transfer learning framework for the detection of fetal cardiac anomalies using ultrasound scan images. The research framework uses four datasets, FetalEcho_V0504, FetalEcho_V0505, FetalEcho_V0506, and FetalEcho_V0507. The first one is the synthetic image dataset developed using GAN. The next three datasets are developed using the image transformation framework where three transformation techniques were developed. The transfer learning framework developed sixteen classifiers using four datasets and four deep learning models. These FCA detection classifiers are demonstrating average precision, recall, accuracy and F1-score of 0.94, 0.93, 0.93 and 0.92 respectively and the AUC of the best classifier identified is 0.99. The significance of individual performance of each classifier is identified using Friedman statistical test and based on the significance the evaluation of classifiers are performed to identify the best classifier.