Scholarlink

<p>Heart disease (HD) or cardiovascular disease (CD) cause about 17.9&#xa0;million deaths annually, or 32% of all deaths worldwide. Mortality rates can be substantially decreased by early HD discovery and adequate treatment, with prompt intervention prior to illness progression increasing treatment success. Regular medical exams and the monitoring of important indicators, such as blood pressure fluctuations, cholesterol levels, diabetes, and obesity, can lead to early detection. In order to detect the existence of HDs at an early stage, this publication presents a cardiovascular disease prediction (HDP) prediction. This research investigated a novel EfficientNetB0 and Fewshot transformer model for HD prediction. The pattern goes through phases of training, evaluation, pattern creation, and data preparation. The accuracy, precision, recall, and F1-rating of the conventional CNN pattern were 86.9%, 84%, 81%, and 83%, respectively. The pattern performs better when the Fewshot transformer technique is used with accuracy, precision, recall, and F1-rating of 85.25%, 90%, 81%, and 85%, respectively. Additionally, superior performance with accuracy, precision, recall, and F1-rating of 86.9%, 87.5%, 87.5%, and 87.5%, respectively, was achieved using the hybrid technique. The precision-recall curve (PRC) and receiver operating characteristics (ROC) area under the curve (AUC) demonstrate how well the suggested hybrid approach performs. These results highlight the efficacy of combining EfficientNetB0 and Fewshot transformer technique for early-stage HDP, providing notable improvements in medical diagnosis.</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

An efficientnet-fewshot transformer driven learning architecture for robust coronary heart disease detection from medical images

  • 期刊论文

  • 发表日期：

    2026年6月25日

查看全文
  • Tejashree Venkatesha,
  • Saravana Kumar Sundararajan

摘要

Heart disease (HD) or cardiovascular disease (CD) cause about 17.9 million deaths annually, or 32% of all deaths worldwide. Mortality rates can be substantially decreased by early HD discovery and adequate treatment, with prompt intervention prior to illness progression increasing treatment success. Regular medical exams and the monitoring of important indicators, such as blood pressure fluctuations, cholesterol levels, diabetes, and obesity, can lead to early detection. In order to detect the existence of HDs at an early stage, this publication presents a cardiovascular disease prediction (HDP) prediction. This research investigated a novel EfficientNetB0 and Fewshot transformer model for HD prediction. The pattern goes through phases of training, evaluation, pattern creation, and data preparation. The accuracy, precision, recall, and F1-rating of the conventional CNN pattern were 86.9%, 84%, 81%, and 83%, respectively. The pattern performs better when the Fewshot transformer technique is used with accuracy, precision, recall, and F1-rating of 85.25%, 90%, 81%, and 85%, respectively. Additionally, superior performance with accuracy, precision, recall, and F1-rating of 86.9%, 87.5%, 87.5%, and 87.5%, respectively, was achieved using the hybrid technique. The precision-recall curve (PRC) and receiver operating characteristics (ROC) area under the curve (AUC) demonstrate how well the suggested hybrid approach performs. These results highlight the efficacy of combining EfficientNetB0 and Fewshot transformer technique for early-stage HDP, providing notable improvements in medical diagnosis.